Pain and Memory: How Chronic Pain Occupies the Working Memory You Need

Introduction

You read the same paragraph four times. You get to the end of it and nothing stayed. Not one clause. You go back to the top, and this time you concentrate, and it happens again.

If your back has hurt every day for two years, you have probably decided this means your memory is failing. That is the natural conclusion. It is also the wrong one.

The paragraph never made it into memory because it never got through the door. Memory formation runs through attention. Attention is limited. And something else is already using it.

Pain is not a background sensation that you slowly stop noticing. It is a signal built to interrupt. That is its entire evolutionary job. A stimulus that announced tissue damage quietly, without demanding that you drop everything and deal with it, would be useless. So pain does not sit in the background politely. It reaches into the same narrow channel your reading was running in, and it takes it [1].

That framing changes what the problem is. It is not a storage failure. Nothing is being deleted. The information simply never got encoded, because the resource encoding depends on was already spoken for. You did not forget the paragraph. You never had it.

This matters at a scale most people underestimate. In the United States alone, national survey data put chronic pain among adults in the tens of millions, with a substantial subset reporting pain severe enough to limit daily activity [2][3]. A meaningful fraction of those people are quietly convinced they are losing their minds.

They are not. What follows is what the research actually found, including the parts that argue with each other, the number everybody quotes without citing, and the one finding that reverses what you have probably been told about your medication.

What Pain Is Actually For

In 1999, Chris Eccleston and Geert Crombez published a paper in Psychological Bulletin that has quietly shaped every serious discussion of this topic since. Its title is the argument: pain demands attention [1].

Their model treats pain as an interrupt. Not as a sensation you feel, but as an event that stops other processing. The reason it works that way is obvious once you say it out loud. An organism that could ignore tissue damage while it finished reading would not last long.

So pain is engineered to win. It has priority in the queue.

The brain has machinery for exactly this kind of arbitration. A network centred on the anterior insula and the anterior cingulate cortex acts as a switch, monitoring incoming signals for anything urgent enough to justify reorganising what you are currently doing [4]. When something clears that bar, the switch flips. Attention moves. Whatever was running gets suspended.

Pain clears that bar constantly. That is the point of it.

The same cingulate region that handles this arbitration also sits at the junction of negative affect, pain and cognitive control, which is not a coincidence [5]. These three things share hardware. When one of them is running hot, the others have less to work with.

Eccleston had already been building the empirical side of this before the model was published. His 1995 work looked at what sustained and shifting attention do in people with chronic pain, and his 1997 paper examined how somatic awareness competes with everything else [6][7]. The pattern in both is the same. Attention is a finite pool, and pain is drawing on it.

None of this is exotic. Attention research outside the pain field arrives at the same architecture from a completely different direction, describing a small number of interacting systems that decide what gets processed and what does not [8]. The prefrontal cortex runs the control side of that arrangement, holding goals in place against distraction [9].

If you have read our piece on how attention gates what your brain keeps, you already know the shape of the argument. Attention is not a nice-to-have for memory. It is the gate. Close the gate and nothing gets in, no matter how good the storage behind it is.

Pain closes the gate. Not completely, and not all the time. But often enough, and for long enough, that it shows up in measurement.

The Number Everyone Quotes and Nobody Sources

Search for chronic pain and memory and you will hit the same statistic within about thirty seconds. Two-thirds of chronic pain patients have measurable cognitive impairment.

It appears on clinic blogs, patient forums and assessment-company marketing pages. Almost none of them say where it came from. Fewer still say what it actually measured.

It came from Bruce Dick and Saifudin Rashiq, writing in Anesthesia and Analgesia in 2007 [10]. They gave people with chronic pain computerised tests of working memory plus a neuropsychological test of attention. Crucially, they tested before and after procedures that produced actual analgesia.

Two-thirds scored in the clinically impaired range on the attentional tasks. That is the number.

Now here is the part that gets dropped every single time, and it is the part that matters. The result held independent of age. It held independent of education level. It held independent of sleep disruption. And it held independent of pain relief.

Read that last one again. The impairment did not simply vanish when the pain did.

Those four controls are not decoration. They are the study pre-emptively knocking down the four explanations a sceptical reader reaches for first. It is not that pain patients are older. It is not that they are less educated. It is not just that they slept badly. And it is not simply a matter of how much it happens to hurt in the moment.

The participants with the worst impairment also had the most trouble maintaining a memory trace, which is the specific mechanism the interruption model predicts. Something is disrupting the holding of information, not the filing of it.

Quote the two-thirds if you like. Just quote the controls with it.

Two Meta-Analyses and One Careful Word

A single study, however well controlled, is a single study. The question is what happens when you pool the field.

Carolyn Berryman and colleagues did that twice. In 2013 they published a systematic review and meta-analysis in Pain, built to Cochrane and PRISMA standards, searching six databases with a research librarian, and pulling in 24 observational studies that compared a chronic pain group against a control group [11]. The target was working memory specifically.

The following year they did the same for executive function, reviewing 25 studies and pooling 22 of them [12].

Their finding, in their own words, was a small to moderate impairment.

Hold on to that phrasing. It is doing more work than it looks like. Small to moderate is not nothing, and it is not catastrophe. It means the average person with chronic pain performs measurably worse on these tasks than the average person without, and that the gap is real but not enormous. It also means that plenty of individuals with chronic pain will test entirely normally, which is exactly what clinicians see.

The broader clinical literature agrees on the direction. Orla Moriarty and colleagues reviewed the clinical and preclinical evidence together in 2011 and reached the same conclusion from a wider base [13]. Earlier neuropsychological work had already flagged the pattern [14], and studies in community-dwelling older adults tied pain to both neuropsychological performance and physical function [15].

More recent syntheses continue to land in the same place [16][17], including cross-sectional work comparing chronic pain patients directly against healthy controls [18].

So there is a consensus. It is just a more modest consensus than the internet suggests.

Which Memory Is Actually Affected

This is where most coverage goes badly wrong, and where the honest answer is genuinely reassuring.

People say "memory". The literature does not. It says working memory, sustained attention, executive function, and the attention-demanding parts of retrieval. It does not say long-term storage.

That distinction is not academic. It is the difference between the fog you have and the disease you are afraid of.

Your childhood is fine. The name of the street you grew up on is fine. What is not fine is holding four things in your head at once while somebody talks at you, which is what working memory under load actually demands.

A doctor's appointment is a working memory task. So is following a recipe you have not made before. So is reading a dense paragraph. Every one of those requires you to hold pieces in place while you assemble them, and holding is the thing pain interferes with.

This is why the fog feels so specific and so humiliating. You can tell someone a detailed story about 1994 and then lose the thread of a sentence you started ten seconds ago. Nothing about that is contradictory. Those are two different systems, and only one of them is under attack.

The Disagreement Nobody Mentions

Here is a fact you will not find on a single page-one search result: the literature is not unanimous.

In 2022, a team led by Sung Eun Jang ran two systematic reviews side by side in Cureus, one on whether pain impairs cognition and one on whether opioids do [19]. For the first question they screened 1,786 articles and found 23 that met their criteria.

Of those 23, sixteen concluded that chronic pain patients showed impaired cognitive function. Six concluded that chronic pain does not worsen cognitive function. One concluded that the answer depends on the patient's baseline cognitive status.

Sixteen to six is a clear majority. It is not unanimity, and pretending otherwise would be dishonest.

Why would six competent studies find nothing? Several reasons, all mundane. Different pain conditions behave differently. Test batteries vary enormously in how much load they impose, and a task that is easy enough will not reveal a small deficit in anybody. Sample sizes in this field are often modest. And the effect being hunted is, in Berryman's pooled estimate, small to moderate, which is precisely the size that appears and disappears depending on how well powered your study is.

That is not a scandal. That is what a real, modest, genuine effect looks like from the inside of a literature.

You should be more suspicious of a field where every study agrees.

The Argument Runs Backwards Too

If pain and thinking really do draw on one shared pool, then the relationship should be symmetric. Spending the resource on a demanding task should leave less of it available for pain.

It does.

Distraction analgesia is a well documented effect, and recent work has shown it to be working-memory-load dependent, with a cortical signature that tracks how hard the task is [20]. Give someone a genuinely difficult mental task and their reported pain drops. Give them an easy one and it does not.

That is a much stronger piece of evidence than it first appears. Any theory can explain a correlation in one direction. A theory that predicts the reverse effect, and then finds it, is doing real work.

The experiment also runs cleanly in the other configuration. Apply thermal pain to healthy volunteers and their n-back performance degrades [21]. No chronic illness involved. No medication. No sleep debt, no depression, no disability, none of the things that follow a person into a pain clinic. Just pain, and worse working memory.

That is about as clean an isolation of the variable as this field gets.

It also explains something people with chronic pain report constantly and rarely get believed about: that they can function fine while genuinely absorbed in something, and fall apart the moment the absorption breaks. That is not inconsistency. That is sustained attention holding the line until it cannot.

A Fog That Turns Up Where You Would Not Expect It

If this really is a property of pain, and not a quirk of one diagnosis, it should show up wherever pain is persistent. Different conditions, different mechanisms, same cognitive fingerprint.

Test that against burning mouth syndrome. It is an uncommon condition, poorly known outside dentistry and neurology, with nothing in common with a herniated disc except that it hurts and does not stop. Researchers looking at cognition in these patients found the same picture and gave their paper a name that says it all: Burning Fog [22].

Migraine gives the same answer. A 2024 review pulled the migraine cognition literature together, and follow-up work has gone after the mechanisms specifically [23][24]. There is even evidence comparing performance between attacks and during them, which is a natural experiment most conditions cannot offer [25].

Irritable bowel syndrome and inflammatory bowel disease, again [26]. Endometriosis-associated chronic pelvic pain, again [27].

Four conditions. Four unrelated parts of the body. Four different disease processes. One cognitive profile.

This is why the field increasingly talks about nociplastic pain as a category in its own right, defined by altered pain processing rather than by ongoing tissue damage [28][29]. Fibromyalgia reviews now treat the cognitive complaint as a core feature rather than an aside [30]. If the cognitive cost travels with the pain rather than with the injury, that is a strong hint about where the cost is being incurred.

Is It Just Depression? The Question Patients Get Asked First

Anyone who has been through a pain clinic knows this question. Sometimes it is asked kindly. Sometimes it is not asked at all, just assumed.

It deserves a real answer, and it has one.

Olga Gelonch and colleagues took 105 women with fibromyalgia, aged 30 to 55, and put them through a full neuropsychological assessment covering attention and executive function, alongside self-report inventories for subjective cognitive complaints, depression, anxiety and pain intensity [31]. The design asks the question directly: are these complaints objective dysfunction, or are they depression wearing a different coat?

The short version is that depression does not account for the objective findings. It contributes. It does not explain them away.

That distinction matters enormously to the person living it, and it is worth being precise about why. Nobody in this literature is claiming depression is irrelevant. Depression and chronic pain are deeply entangled, and have been described in mediation terms since at least the 1980s [32]. Mood shapes what you encode and what you retrieve, which is a whole subject in itself and one we have written about in how emotion shapes memory.

The claim is narrower and stronger. When you measure objective cognitive performance and statistically account for depressive symptoms, a deficit is still sitting there.

There is a second layer to this that is worth naming. People with chronic pain also develop measurable attentional biases toward pain-related information, which is not the same thing as cognitive impairment but interacts with it [33][34][35]. Hypervigilance is well documented, in fibromyalgia particularly, and there are validated instruments for measuring it [36][37][38].

Vigilance costs something. A system that keeps checking on a threat has less capacity left for anything else. Catastrophizing, which is partly a vigilance phenomenon, tracks both brain responses to pain and cognitive outcomes [39][40].

So the honest answer to "is it just depression" is no, with an asterisk that is not a dismissal. Mood, vigilance and pain are braided together. Untangling them completely may not even be the right goal. But the objective deficit survives the untangling that has been done.

Sleep and Stress Are Real. They Are Not the Whole Answer.

The other two suspects deserve the same treatment.

Sleep and pain have one of the nastiest feedback loops in medicine. Pain wrecks sleep. Poor sleep amplifies pain. The relationship has been documented longitudinally and experimentally for two decades [41][42], and recent observational work continues to link sleep quality directly to cognitive performance in chronic musculoskeletal pain [43]. Central sensitisation severity tracks sleep disturbance too [44].

Anyone who has had a bad night knows what it does to concentration. Our piece on what the sleeping brain does with the day's memories covers the consolidation side of that.

Stress is the third suspect, and there is a particularly careful study on it. Henrik Jacobsen and colleagues collected hair samples from 122 people, giving them a measure of chronic cortisol exposure over months rather than a single stressed morning [45]. The group included 40 people with fibromyalgia, 24 with peripheral neuropathic pain, and 58 matched healthy controls. Of those, 84 also completed detailed executive function testing.

Chronic stress had its own specific effect on working memory. Not a vague association with feeling bad. A measurable, targeted hit.

That fits a much larger literature on stress and executive function, where meta-analysis shows acute stress reliably degrades core executive processes [46]. We have gone into the cortisol mechanism separately in what stress hormones do to memory.

Now put all the suspects in one place.

Read that table from the top down and a pattern appears. Every suspect has a case against it. None of them clears the pain signal itself, because the 2007 controls already ruled out the two most obvious alternatives.

Sleep and stress are contributors. They are not substitutes for the explanation.

Then It Must Be the Painkillers

This is the point where nearly every article you have read reaches for the obvious villain.

The reasoning is intuitive. Opioids are sedating. Sedation impairs cognition. Therefore the fog is pharmacological, and the fix is to take less.

The evidence does not support that story, and this is the single most useful thing in this entire article.

In 2021, a team led by Jane Akhurst published a systematic review and meta-analysis in Pain Medicine, searching EMBASE, Medline and PsycINFO and pooling 17 studies across five cognitive domains: motor performance, attention, working memory, executive functions and memory [47]. What makes this study unusually good is the control groups. It did not just compare opioid users against healthy people. It included comparisons against chronic pain patients who were not taking opioids, and against the same patients before they started.

Against healthy controls, opioid users showed deficits in attention and memory. That is the finding that circulates.

Against chronic pain patients who were not on opioids, the difference largely disappeared.

And compared with their own opioid-free baseline, patients tended to show improvement in attention and working memory after starting treatment.

Sit with that for a moment. Treating the pain made the thinking better, not worse.

Jang's parallel review reached a compatible conclusion from a separate evidence base, screening 584 articles on the opioid question and finding 18 eligible [19]. The practical upshot in both is the same: under-treated pain is itself cognitively expensive, and concern about cognition is a weak reason to leave significant pain untreated.

The counterweight is real and should be stated. The Cochrane review of adverse events from medium and long-term opioid use in chronic non-cancer pain documents genuine harms, and none of the above says opioids are consequence-free [48]. Dose matters. Escalation matters. Individual response varies enormously.

But the specific claim that your medication is the reason you cannot follow a conversation is not what the comparison studies show.

Nothing in this article is advice about anybody's prescription. What to take and how much of it is a conversation between a person and their clinician, informed by far more than one variable. The point here is narrower: if you have been quietly blaming the pills, the evidence says the pain is the better suspect.

What the Brain Looks Like While This Is Happening

Human imaging has been circling this for nearly two decades.

The landmark result came from Marwan Baliki and colleagues in 2008, under a title that did not hedge: chronic pain hurts the brain, disrupting the default-mode network [49]. The default-mode network is the pattern of activity your brain settles into when it is not doing a task. It has to deactivate properly for you to engage with something external. In chronic pain, that deactivation is disturbed.

That is a mechanistic description of not being able to settle into a task.

The same group had earlier shown that chronic pain involves brain activity distinct from acute pain, sitting in emotional rather than sensory circuitry [50], and later work found this network reorganisation across multiple different chronic pain conditions [51]. The prefrontal cortex has since been characterised as a central player in pain processing rather than a bystander [52]. In fibromyalgia, intrinsic connectivity tracks reported pain intensity [53]. And chronic pain patients perform poorly on emotional decision-making tasks that depend on prefrontal function [54].

Here is where it gets more complicated, and where a popular simplification breaks.

The story that circulates is that chronic pain shrinks brain tissue and that is why you forget. Structural change is real and repeatedly documented. But a 2025 NeuroImage study went looking for the link between working memory impairment and grey matter volume change in women with fibromyalgia, and found the two decoupled [55].

Both things change. The change in one does not neatly explain the change in the other.

That is an inconvenient result and it belongs in any honest account. Structure is not destiny, and a simple volume-loss story does not survive contact with the data.

Animal work fills in some of what human imaging cannot resolve. Rodent studies show morphological and functional reorganisation of the medial prefrontal cortex in neuropathic pain [56], with excitatory inputs to that region rearranging over the course of chronic pain [57]. Those are rodents. The caveat is not a formality.

Down to the Synapse and Into a Mouse

The mechanism work has moved quickly in the last two years, and none of it has reached a page that a patient would ever read.

In 2024, a group at Xuzhou Medical University led by Mengqiao Cui, with fifteen authors on the paper, published a result in eLife tying pain-related memory impairment to a specific signalling pathway in the dentate gyrus of the hippocampus [58]. The pathway involves sphingosine 1-phosphate and its receptor S1PR1, which helps maintain the actin scaffolding inside dendritic spines. Lose that maintenance and the physical connections between neurons weaken.

The hippocampus is where the brain decides what gets kept, and the dentate gyrus is one of its most plastic regions. Elsewhere in the same structure, the coordinated bursts known as sharp wave ripples are treated as a biological marker of episodic memory and planning [59]. Degrading spine structure in a region feeding that machinery is a plausible route from persistent pain to memory failure.

This is mouse work. That is not a small caveat and I am not going to bury it.

eLife published an accompanying commentary by Suelen Pereira, Ivan Tomsic, Robson da Costa and Mychael Lourenço that says so plainly and lists the open questions [60]. Does S1PR1 influence memory types beyond spatial memory? Do these mechanisms apply to other neurological conditions? And, in their own framing, how well do results in mice translate to humans?

That is a research field being honest about its own limits in print. It deserves to be quoted rather than quietly skipped.

Other preclinical results point in compatible directions. Chronic pain has been shown to produce tau-mediated hippocampal pathology and memory deficits in animal models, which is a striking bridge between the pain literature and the dementia literature [61]. Impaired WNT3 and IGF-1 signalling in the dorsal dentate gyrus has been implicated [62], as has glia-derived adenosine in the ventral hippocampus [63].

Neuroinflammation runs underneath much of this. The anterior cingulate cortex undergoes neuroinflammatory remodelling in chronic pain states [64], and extracellular RNA signalling through TLR3 has been linked to cognitive impairment after chronic neuropathic pain [65]. Broader reviews now attempt to pull the clinical picture and these mechanisms into one frame [66][67].

Every one of those mechanistic findings is preclinical. Not one of them has been demonstrated in a human being. They are the best current guesses at what is physically happening, and guesses at that level have a long history of not surviving translation.

What they do establish is that there is something to find. This is not a psychological artefact in search of a biology.

The Long Shadow

Everything so far is about right now. There is a separate question about the long run, and it has become much better answered in the last three years.

In 2023, Wenhui Zhao and colleagues used the UK Biobank to ask whether pain at multiple body sites carries more cognitive risk than pain at one [68]. They analysed dementia risk in 354,943 people, then examined cognition and brain structure in a subset of 19,116. More pain sites meant more risk. Not a subtle gradient either.

Then in 2026 a meta-analysis in Translational Psychiatry pooled the longitudinal evidence properly: 28 cohorts, 7,914,407 participants, searched through to January 2025 [69]. Chronic pain was associated with a higher risk of cognitive impairment, with a pooled adjusted odds ratio of 1.30 and a confidence interval of 1.14 to 1.47.

The detail underneath that headline matters more than the headline. The effect was driven by dementia diagnoses, with a pooled odds ratio of 1.43, rather than by global cognitive performance scores.

That is a strange and important pattern. Diagnoses go up more than test scores do.

There are at least two readings and I am not going to pretend to adjudicate between them. It could mean the association is genuinely about eventual neurodegeneration. It could also mean that people in chronic pain come to more medical attention and get diagnosed more readily. Observational cohorts cannot separate those, no matter how many million people are in them.

Association is not causation. Nearly eight million participants does not change that. It just makes the association very hard to dismiss.

Prospective work is starting to close in on the mechanism side. A 2025 UK Biobank study looked at nociplastic pain severity, scored on the fibromyalgia index, against a latent executive function measure derived by confirmatory factor analysis, with baseline cognitive testing in 2017 to 2020 and follow-up in 2021 to 2022 [70]. That design lets you watch decline happen rather than infer it. Other recent work suggests pain interference, meaning how much pain gets in the way, predicts cognitive outcomes better than the mere presence of pain [71], which is exactly what an interruption model would predict. In older adults specifically, the relationship has been reviewed in its own right [72].

If you take one thing from this section, take the distinction. A measurable working memory cost today and an elevated statistical risk over decades are two different claims, supported by two different kinds of evidence, and they should never be collapsed into "chronic pain gives you dementia".

Fourteen People Who Got Some of It Back

Almost every article on this subject ends with coping tips. Very few ask the question that actually matters: does any of it come back?

There is one study that gets closest to an answer, and it is small enough that I am going to give you the number before the finding.

David Seminowicz and colleagues scanned 18 patients with chronic low back pain, then treated them with either spine surgery or facet joint injections, then rescanned the 14 who returned six months later [73]. They also scanned 16 healthy controls, of whom 10 came back at six months for comparison.

Before treatment, the patients had a thinner left dorsolateral prefrontal cortex than controls. After successful treatment, cortical thickness in that region increased.

Fourteen people. Six months. One region.

That is not a promise and I am not going to sell it as one. It is a small longitudinal imaging study, and small longitudinal imaging studies have a poor track record of replicating exactly. But it is the most direct evidence available that the structural changes seen in chronic pain are not necessarily permanent, and that they track treatment outcome rather than simply accumulating with time.

Interventional work is beginning to test the cognitive side directly. A randomised trial has evaluated pain neuroscience education against executive function in women with fibromyalgia [74], and a double-blind placebo-controlled trial has tested attentional bias modification delivered over the web [75]. These are early and their results should be read as early.

The direction of travel is what is interesting. The field has stopped treating the cognitive cost as an unavoidable side effect and started treating it as something with a mechanism, and therefore as something that might have a target.

How the Evidence Arrived

Laid out end to end, the history of this question is shorter than you would expect and most of the mechanism sits in the last three years.

Notice how the mechanism arrives last. That is normal. Fields usually measure an effect for decades before anyone can say what is physically causing it, and the measurement here has been solid since 2007 while the biology only started resolving in 2024.

Notice also how much of the recent work is in animals. The 2024 and 2025 mechanism papers are almost all rodent. The 2023 and 2026 epidemiology is enormous and human but observational. What is thin in the middle is human mechanistic work, and that gap is the honest description of where this field currently stands.

What This Actually Means

If you have chronic pain and you have been frightened by your own memory, the evidence supports a fairly specific reassurance, and it is worth stating carefully rather than as a list of comforting bullet points.

Start with the fact that the problem is real. It has been measured repeatedly, by different groups, and pooled across dozens of studies, so it is neither imagined nor a personal failing. What those same pooled analyses also say is that the effect is moderate. Berryman's phrase was small to moderate, and that phrase deserves to travel with every claim in this area, because plenty of people with chronic pain test entirely normally.

More usefully, the damage is narrow. Working memory, sustained attention and effortful retrieval take the hit while long-term storage largely holds, which is a fundamentally different picture from progressive memory disease. That single distinction is probably the most valuable thing on this page.

Three obvious alternative explanations have each been tested and none of them replaces it. Tiredness does not, because the 2007 controls covered sleep disruption. Low mood does not, because the fibromyalgia work measured depression directly and the deficit survived it. And your medication probably does not either, because when researchers compared people on opioids against people carrying the same pain without them, the difference largely went away.

There is even a small signal that none of it is fixed forever, coming from fourteen people whose prefrontal cortex thickened again once their pain was treated.

What none of this is, is instruction. This article makes no recommendation about anybody's treatment, and questions about your own care belong with the clinician who knows your case.

The reframe is the useful part. You are not failing to remember. You are being interrupted, constantly, by a signal built specifically to interrupt you, and that signal is drawing down the same limited resource that reading and listening and holding a thought require.

You did not forget the paragraph.

You never got the chance to learn it.