Sensorimotor Memory Research Articles

Sensorimotor Memory Biases Weight Perception During Object Lifting.

When lifting an object, the brain uses visual cues and an internal object representation to predict its weight and scale fingertip forces accordingly. Once available, tactile information is rapidly integrated to update the weight prediction and refine the internal object representation. If visual cues cannot be used to predict weight, force planning relies on implicit knowledge acquired from recent lifting experience, termed sensorimotor memory. Here, we investigated whether perception of weight is similarly biased according to previous lifting experience and how this is related to force scaling. Participants grasped and lifted series of light or heavy objects in a semi-randomized order and estimated their weights. As expected, we found that forces were scaled based on previous lifts (sensorimotor memory) and these effects increased depending on the length of recent lifting experience. Importantly, perceptual weight estimates were also influenced by the preceding lift, resulting in lower estimations after a heavy lift compared to a light one. In addition, weight estimations were negatively correlated with the magnitude of planned force parameters. This perceptual bias was only found if the current lift was light, but not heavy since the magnitude of sensorimotor memory effects had, according to Weber’s law, relatively less impact on heavy compared to light objects. A control experiment tested the importance of active lifting in mediating these perceptual changes and showed that when weights are passively applied on the hand, no effect of previous sensory experience is found on perception. These results highlight how fast learning of novel object lifting dynamics can shape weight perception and demonstrate a tight link between action planning and perception control. If predictive force scaling and actual object weight do not match, the online motor corrections, rapidly implemented to downscale forces, will also downscale weight estimation in a proportional manner.

Infants' prospective control during object manipulation in an uncertain environment.

This study investigates how infants use visual and sensorimotor information to prospectively control their actions. We gave 14-month-olds two objects of different weight and observed how high they were lifted, using a Qualisys Motion Capture System. In one condition, the two objects were visually distinct (different color condition) in another they were visually identical (same color condition). Lifting amplitudes of the first movement unit were analyzed in order to assess prospective control. Results demonstrate that infants lifted a light object higher than a heavy object, especially when vision could be used to assess weight (different color condition). When being confronted with two visually identical objects of different weight (same color condition), infants showed a different lifting pattern than what could be observed in the different color condition, expressed by a significant interaction effect between object weight and color condition on lifting amplitude. These results indicate that (a) visual information about object weight can be used to prospectively control lifting actions and that (b) infants are able to prospectively control their lifting actions even without visual information about object weight. We argue that infants, in the absence of reliable visual information about object weight, heighten their dependence on non-visual information (tactile, sensorimotor memory) in order to estimate weight and pre-adjust their lifting actions in a prospective manner.

The influence of visual feedback from the recent past on the programming of grip aperture is grasp-specific, shared between hands, and mediated by sensorimotor memory not task set

Goal-directed movements, such as reaching out to grasp an object, are necessarily constrained by the spatial properties of the target such as its size, shape, and position. For example, during a reach-to-grasp movement, the peak width of the aperture formed by the thumb and fingers in flight (peak grip aperture, PGA) is linearly related to the target’s size. Suppressing vision throughout the movement (visual open loop) has a small though significant effect on this relationship. Visual open loop conditions also produce a large increase in the PGA compared to when vision is available throughout the movement (visual closed loop). Curiously, this differential effect of the availability of visual feedback is influenced by the presentation order: the difference in PGA between closed- and open-loop trials is smaller when these trials are intermixed (an effect we have called ‘homogenization’). Thus, grasping movements are affected not only by the availability of visual feedback (closed loop or open loop) but also by what happened on the previous trial. It is not clear, however, whether this carry-over effect is mediated through motor (or sensorimotor) memory or through the interference of different task sets for closed-loop and open-loop feedback that determine when the movements are fully specified. We reasoned that sensorimotor memory, but not a task set for closed and open loop feedback, would be specific to the type of response. We tested this prediction in a condition in which pointing to targets was alternated with grasping those same targets. Critically, in this condition, when pointing was performed in open loop, grasping was always performed in closed loop (and vice versa). Despite the fact that closed- and open-loop trials were alternating in this condition, we found no evidence for homogenization of the PGA. Homogenization did occur, however, in a follow-up experiment in which grasping movements and visual feedback were alternated between the left and the right hand, indicating that sensorimotor (or motor) memory can operate both within and between hands when the response type is kept the same. In a final experiment, we ruled out the possibility that simply alternating the hand used to perform the grasp interferes with motor or sensorimotor memory. We did this by showing that when the hand was alternated within a block of exclusively closed- or open-loop trials, homogenization of the PGA did not occur. Taken together, the results suggest that (1) interference from simply switching between task sets for closed or open-loop feedback or from switching between the hands cannot account homogenization in the PGA and that (2) the programming and execution of grasps can borrow not only from grasping movements executed in the past by the same hand, but also from grasping movements executed with the other hand.

An fMRI study of differences in brain activity among elite, expert, and novice archers at the moment of optimal aiming.

We investigated brain activity in elite, expert, and novice archers during a simulated archery aiming task to determine whether neural correlates of performance differ by skill level. Success in shooting sports depends on complex mental routines just before the shot, when the brain prepares to execute the movement. During functional magnetic resonance imaging, 40 elite, expert, or novice archers aimed at a simulated 70-meter-distant target and pushed a button when they mentally released the bowstring. At the moment of optimal aiming, the elite and expert archers relied primarily on a dorsal pathway, with greatest activity in the occipital lobe, temporoparietal lobe, and dorsolateral pre-motor cortex. The elites showed activity in the supplementary motor area, temporoparietal area, and cerebellar dentate, while the experts showed activity only in the superior frontal area. The novices showed concurrent activity in not only the dorsolateral pre-motor cortex but also the ventral pathways linked to the ventrolateral pre-motor cortex. The novices exhibited broad activity in the superior frontal area, inferior frontal area, ventral prefrontal cortex, primary motor cortex, superior parietal lobule, and primary somatosensory cortex. The more localized neural activity of elite and expert archers than novices permits greater efficiency in the complex processes subserved by these regions. The elite group's high activity in the cerebellar dentate indicates that the cerebellum is involved in automating simultaneous movements by integrating the sensorimotor memory enabled by greater expertise in self-paced aiming tasks. A companion article comments on and generalizes our findings.

Zeta inhibitory peptide (ZIP) erases long-term memories in a cockroach

Recent efforts to identify the molecules that are involved in the maintenance of long-term memories in mammals have focused attention on atypical isoforms of protein kinase C (PKC). Inhibition of these kinases by either the general PKC inhibitor, chelerythrine, or the more specific inhibitor, zeta inhibitory peptide (ZIP), can abolish both long-term potentiation in the hippocampus and as well as spatial, fear, appetitive, and sensorimotor memories. These inhibitors can also abolish long-term facilitation and long-term sensitization in the mollusk Aplysia californica. We have extended these results to an insect, the cockroach Leucophaea maderae. We show that systemic injections of either chelerythrine or ZIP erase long-term olfactory memories in the cockroach, but have no effect on memory acquisition during conditioning. We also show that inhibition of either protein kinase A (PKA) or protein synthesis can block memory acquisition but neither has an effect on the memory once it is formed. The results suggest that sustaining memories in insects requires the persistent activity of one or more isoforms of PKC and point to a strong evolutionary conservation of the molecular mechanisms that underlie the persistence of long-term memories in the central nervous system.

Retention and interference of learned dexterous manipulation: interaction between multiple sensorimotor processes.

An object can be used in multiple contexts, each requiring different hand actions. How the central nervous system builds and maintains memory of such dexterous manipulations remains unclear. We conducted experiments in which human subjects had to learn and recall manipulations performed in two contexts, A and B. Both contexts involved lifting the same L-shaped object whose geometry cued its asymmetrical mass distribution. Correct performance required producing a torque on the vertical handle at object lift onset to prevent it from tilting. The torque direction depended on the context, i.e., object orientation, which was changed by 180° object rotation about a vertical axis. With an A1B1A2 context switching paradigm, subjects learned A1 in the first block of eight trials as indicated by a torque approaching the required one. However, subjects made large errors in anticipating the required torque when switching to B1 immediately after A1 (negative transfer), as well as when they had to recall A1 when switching to A2 after learning B through another block of eight lifts (retrieval interference). Classic sensorimotor learning theories attribute such interferences to multi-rate, multi-state error-driven updates of internal models. However, by systematically changing the interblock break duration and within-block number of trials, our results suggest an alternative explanation underlying interference and retention of dexterous manipulation. Specifically, we identified and quantified through a novel computational model the nonlinear interaction between two sensorimotor mechanisms: a short-lived, context-independent, use-dependent sensorimotor memory and a context-sensitive, error-based learning process.

Representation of Object Weight in Human Ventral Visual Cortex

Skilled manipulation requires the ability to predict the weights of viewed objects based on learned associations linking object weight to object visual appearance. However, the neural mechanisms involved in extracting weight information from viewed object properties are unknown. Given that ventral visual pathway areas represent a wide variety of object features, one intriguing but as yet untested possibility is that these areas also represent object weight, a nonvisual motor-relevant object property. Here, using event-related fMRI and pattern classification techniques, we tested the novel hypothesis that object-sensitive regions in occipitotemporal cortex (OTC), in addition to traditional motor-related brain areas, represent object weight when preparing to lift that object. In two studies, the same participants prepared and then executed lifting actions with objects of varying weight. In the first study, we show that when lifting visually identical objects, where predicted weight is based solely on sensorimotor memory, weight is represented in object-sensitive OTC. In the second study, we show that when object weight is associated with a particular surface texture, that texture-sensitive OTC areas also come to represent object weight. Notably, these texture-sensitive areas failed to carry information about weight in the first study, when object surface properties did not specify weight. Our results indicate that the integration of visual and motor-relevant object information occurs at the level of single OTC areas and provide evidence that the ventral visual pathway is actively and flexibly engaged in processing object weight, an object property critical for action planning and control.

Corticospinal excitability underlying digit force planning for grasping in humans

Control of digit forces for grasping relies on sensorimotor memory gained from prior experience with the same or similar objects and on online sensory feedback. However, little is known about neural mechanisms underlying digit force planning. We addressed this question by quantifying the temporal evolution of corticospinal excitability (CSE) using single-pulse transcranial magnetic stimulation (TMS) during two reach-to-grasp tasks. These tasks differed in terms of the magnitude of force exerted on the same points on the object to isolate digit force planning from reach and grasp planning. We also addressed the role of intracortical circuitry within primary motor cortex (M1) by quantifying the balance between short intracortical inhibition and facilitation using paired-pulse TMS on the same tasks. Eighteen right-handed subjects were visually cued to plan digit placement at predetermined locations on the object and subsequently to exert either negligible force ("low-force" task, LF) or 10% of their maximum pinch force ("high-force" task, HF) on the object. We found that the HF task elicited significantly smaller CSE than the LF task, but only when the TMS pulse coincided with the signal to initiate the reach. This force planning-related CSE modulation was specific to the muscles involved in the performance of both tasks. Interestingly, digit force planning did not result in modulation of M1 intracortical inhibitory and facilitatory circuitry. Our findings suggest that planning of digit forces reflected by CSE modulation starts well before object contact and appears to be driven by inputs from frontoparietal areas other than M1.

Multilevel control of run orientation in Drosophila larval chemotaxis

Chemotaxis is a powerful paradigm to study how orientation behavior is driven by sensory stimulation. Drosophila larvae navigate odor gradients by controlling the duration of their runs and the direction of their turns. Straight runs and wide-amplitude turns represent two extremes of a behavioral continuum. Here we establish that, on average, runs curl toward the direction of higher odor concentrations. We find that the orientation and strength of the local odor gradient perpendicular to the direction of motion modulates the orientation of individual runs in a gradual manner. We discuss how this error-correction mechanism, called weathervaning, contributes to larval chemotaxis. We use larvae with a genetically modified olfactory system to demonstrate that unilateral function restricted to a single olfactory sensory neuron (OSN) is sufficient to direct weathervaning. Our finding that bilateral sensing is not necessary to control weathervaning highlights the role of temporal sampling. A correlational analysis between sensory inputs and behavioral outputs suggests that weathervaning results from low-amplitude head casts implemented without interruption of the run. In addition, we report the involvement of a sensorimotor memory arising from previous reorientation events. Together, our results indicate that larval chemotaxis combines concurrent orientation strategies that involve complex computations on different timescales.

Visuo-proprioceptive interactions during adaptation of the human reach.

We examined whether visual and proprioceptive estimates of transient (midreach) target capture errors contribute to motor adaptation according to the probabilistic rules of information integration used for perception. Healthy adult humans grasped and moved a robotic handle between targets in the horizontal plane while the robot generated springlike loads that varied unpredictably from trial to trial. For some trials, a visual cursor faithfully tracked hand motion. In others, the handle's position was locked and subjects viewed motion of a point-mass cursor driven by hand forces. In yet other trials, cursor feedback was dissociated from hand motion or altogether eliminated. We used time- and frequency-domain analyses to characterize how sensorimotor memories influence performance on subsequent reaches. When the senses were used separately, subjects were better at rejecting physical disturbances applied to the hand than virtual disturbances applied to the cursor. In part, this observation reflected differences in how participants used sensorimotor memories to adapt to perturbations when performance feedback was limited to only proprioceptive or visual information channels. When both vision and proprioception were available to guide movement, subjects processed memories in a manner indistinguishable from the vision-only condition, regardless of whether the cursor tracked the hand faithfully or whether we experimentally dissociated motions of the hand and cursor. This was true even though, on average, perceptual uncertainty in the proprioceptive estimation of movement extent exceeded that of visual estimation by just 47%. In contrast to perceptual tasks wherein vision and proprioception both contribute to an optimal estimate of limb state, our findings support a switched-input, multisensory model of predictive load compensation wherein visual feedback of transient performance errors overwhelmingly dominates proprioception in determining adaptive reach performance.

Effects of visual cues of object density on perception and anticipatory control of dexterous manipulation.

Anticipatory force planning during grasping is based on visual cues about the object’s physical properties and sensorimotor memories of previous actions with grasped objects. Vision can be used to estimate object mass based on the object size to identify and recall sensorimotor memories of previously manipulated objects. It is not known whether subjects can use density cues to identify the object’s center of mass (CM) and create compensatory moments in an anticipatory fashion during initial object lifts to prevent tilt. We asked subjects (n = 8) to estimate CM location of visually symmetric objects of uniform densities (plastic or brass, symmetric CM) and non-uniform densities (mixture of plastic and brass, asymmetric CM). We then asked whether subjects can use density cues to scale fingertip forces when lifting the visually symmetric objects of uniform and non-uniform densities. Subjects were able to accurately estimate an object’s center of mass based on visual density cues. When the mass distribution was uniform, subjects could scale their fingertip forces in an anticipatory fashion based on the estimation. However, despite their ability to explicitly estimate CM location when object density was non-uniform, subjects were unable to scale their fingertip forces to create a compensatory moment and prevent tilt on initial lifts. Hefting object parts in the hand before the experiment did not affect this ability. This suggests a dichotomy between the ability to accurately identify the object’s CM location for objects with non-uniform density cues and the ability to utilize this information to correctly scale their fingertip forces. These results are discussed in the context of possible neural mechanisms underlying sensorimotor integration linking visual cues and anticipatory control of grasping.

Brainspotting: Recruiting the midbrain for accessing and healing sensorimotor memories of traumatic activation

Brainspotting is a psychotherapy based in the observation that the body activation experienced when describing a traumatic event has a resonating spot in the visual field. Holding the attention on that Brainspot allows processing of the traumatic event to flow until the body activation has cleared. This is facilitated by a therapist focused on the client and monitoring with attunement. We set out testable hypotheses for this clinical innovation in the treatment of the residues of traumatic experiences. The primary hypothesis is that focusing on the Brainspot engages a retinocollicular pathway to the medial pulvinar, the anterior and posterior cingulate cortices, and the intraparietal sulcus, which has connectivity with the insula. While the linkage of memory, emotion, and body sensation may require the parietal and frontal interconnections – and resolution in the prefrontal cortex – we suggest that the capacity for healing of the altered feeling about the self is occurring in the midbrain at the level of the superior colliculi and the periaqueductal gray.

Grasping uncertainty: effects of sensorimotor memories on high-level planning of dexterous manipulation

For successful object manipulation, the central nervous system must appropriately coordinate digit placement and force distribution. It is known that digit force planning is significantly influenced by previous manipulations even when object properties cannot be predicted on a trial-to-trial basis. We sought to determine whether this effect extends beyond force control to the coordination of digit placement and force. Subjects grasped and lifted an object whose center of mass (CM) was changed unpredictably across trials. Grasp planning was quantified by measuring the torque generated on the object at lift onset. We found that both digit placement and force were systematically affected by the CM experienced on the previous trial. Additionally, the negative covariation between digit forces and positions typically found for predictable CM presentations was also found for unpredictable CM trials. A follow-up experiment revealed that these effects were not dependent on visual feedback of object roll during object lift on the previous trial. We conclude that somatosensory feedback from previous grasp experience alone can affect high-level grasp planning by constraining the relation between digit force and position even when the task behavioral consequences cannot be reliably predicted. As learning of manipulations often involves interactions with objects in novel environments, the present findings are an important step to understanding the control strategies associated with the integration of sensorimotor memories and motor planning.

Tool-use and the left hemisphere: What is lost in ideomotor apraxia?

Impaired tool related action in ideomotor apraxia is normally ascribed to loss of sensorimotor memories for habitual actions (engrams), but this account has not been tested against a hypothesis of a general deficit in representation of hand-object spatial relationships. Rapid reaching for familiar tools was compared with reaching for abstract objects in apraxic patients (N=9) and in a control group with right hemisphere posterior stroke. The apraxic patients alone showed an impairment in rotating the wrist to correctly grasp an inverted tool but not when inverting the hand to avoid a barrier and grasp an abstract object, and the severity of the impairment in tool reaching correlated with pantomime of tool-use. A second experiment with two apraxic patients tested whether barrier avoidance was simply less spatially demanding than reaching for a tool. However, the patient with damage limited to the inferior parietal lobe still showed a selective problem for tools. These results demonstrate that some apraxic patients are selectively impaired in their interaction with familiar tools, and this cannot be explained by the demands of the task on postural or spatial representation. However, traditional engram theory cannot account for associated problems with imitation of novel actions nor the absence of any correlated deficit in recognition of the methods of grasp of common tools. A revised theory is presented which follows the dorsal and ventral streams model (Milner & Goodale, 2008) and proposes preservation of motor control by the dorsal stream but impaired modulating input to it from the conceptual systems of the left temporal lobe.

Context-Dependent Learning Interferes with Visuomotor Transformations for Manipulation Planning

How the CNS transforms visual information of object properties into motor commands for manipulation is not well understood. We designed novel apparatus and protocols in which human subjects had to learn manipulations in two different contexts. The first task involved manipulating a U-shaped object that can afford two actions by grasping different parts of the same object. The second task involved manipulating two L-shaped objects that were posed at different orientations. In both experiments, subjects learned the manipulation over consecutive trials in one context before switching to a different context. For both objects and tasks, the visual geometric cues were effective in eliciting anticipatory control with little error at the beginning of learning of the first context. However, subjects failed to use the visual information to the same extent when switching to the second context as sensorimotor memory built through eight consecutive repetitions in the first context exerted a strong interference on subjects' ability to use visual cues again when the context changed. A follow-up experiment where subjects were exposed to a pseudorandom sequence of context switches with the U-shaped object revealed that the interference caused by the preceding context persisted even when subjects switched context after only one trial. Our results suggest that learning generalization of dexterous manipulation is fundamentally limited by context-specific learning of motor actions and competition between vision-based motor planning and sensorimotor memory.

Material evidence: interaction of well-learned priors and sensorimotor memory when lifting objects

Skilled object lifting requires the prediction of object weight. When lifting new objects, such prediction is based on well-learned size-weight and material-density correlations, or priors. However, if the prediction is erroneous, people quickly learn the weight of the particular object and can use this knowledge, referred to as sensorimotor memory, when lifting the object again. In the present study, we explored how sensorimotor memory, gained when lifting a given object, interacts with well-learned material-density priors when predicting the weight of a larger but otherwise similar-looking object. Different groups of participants 1st lifted 1 of 4 small objects 10 times. These included a pair of wood-filled objects and a pair of brass-filled objects where 1 of each pair was covered in a wood veneer and the other was covered in a brass veneer. All groups then lifted a larger, brass-filled object with the same covering as the small object they had lifted. For each lift, we determined the initial peak rate of change of vertical load-force rate and the load-phase duration, which provide estimates of predicted object weight. Analysis of the 10th lift of the small cube revealed no effects of surface material, indicating participants learned the appropriate forces required to lift the small cube regardless of object appearance. However, both surface material and core material of the small cube affected the 1st lift of the large block. We conclude that sensorimotor memory related to object density can contribute to weight prediction when lifting novel objects but also that long-term priors related to material properties can influence the prediction.

Spatiotemporal dynamics of brain activity during the transition from visually guided to memory-guided force control

It is well established that the prefrontal cortex is involved during memory-guided tasks whereas visually guided tasks are controlled in part by a frontal-parietal network. However, the nature of the transition from visually guided to memory-guided force control is not as well established. As such, this study examines the spatiotemporal pattern of brain activity that occurs during the transition from visually guided to memory-guided force control. We measured 128-channel scalp electroencephalography (EEG) in healthy individuals while they performed a grip force task. After visual feedback was removed, the first significant change in event-related activity occurred in the left central region by 300 ms, followed by changes in prefrontal cortex by 400 ms. Low-resolution electromagnetic tomography (LORETA) was used to localize the strongest activity to the left ventral premotor cortex and ventral prefrontal cortex. A second experiment altered visual feedback gain but did not require memory. In contrast to memory-guided force control, altering visual feedback gain did not lead to early changes in the left central and midline prefrontal regions. Decreasing the spatial amplitude of visual feedback did lead to changes in the midline central region by 300 ms, followed by changes in occipital activity by 400 ms. The findings show that subjects rely on sensorimotor memory processes involving left ventral premotor cortex and ventral prefrontal cortex after the immediate transition from visually guided to memory-guided force control.

Contribution of Cerebellar Sensorimotor Adaptation to Hippocampal Spatial Memory

Complementing its primary role in motor control, cerebellar learning has also a bottom-up influence on cognitive functions, where high-level representations build up from elementary sensorimotor memories. In this paper we examine the cerebellar contribution to both procedural and declarative components of spatial cognition. To do so, we model a functional interplay between the cerebellum and the hippocampal formation during goal-oriented navigation. We reinterpret and complete existing genetic behavioural observations by means of quantitative accounts that cross-link synaptic plasticity mechanisms, single cell and population coding properties, and behavioural responses. In contrast to earlier hypotheses positing only a purely procedural impact of cerebellar adaptation deficits, our results suggest a cerebellar involvement in high-level aspects of behaviour. In particular, we propose that cerebellar learning mechanisms may influence hippocampal place fields, by contributing to the path integration process. Our simulations predict differences in place-cell discharge properties between normal mice and L7-PKCI mutant mice lacking long-term depression at cerebellar parallel fibre-Purkinje cell synapses. On the behavioural level, these results suggest that, by influencing the accuracy of hippocampal spatial codes, cerebellar deficits may impact the exploration-exploitation balance during spatial navigation.

A Model for Mild Traumatic Brain Injury that Induces Limited Transient Memory Impairment and Increased Levels of Axon Related Serum Biomarkers

Mild traumatic brain injury (mTBI) is one of the most common neuronal insults and can lead to long-term disabilities. mTBI occurs when the head is exposed to a rapid acceleration-deceleration movement triggering axonal injuries. Our limited understanding of the underlying pathological changes makes it difficult to predict the outcome of mTBI. In this study we used a scalable rat model for rotational acceleration TBI, previously characterized for the threshold of axonal pathology. We have analyzed whether a TBI just above the defined threshold would induce any detectable behavioral changes and/or changes in serum biomarkers. The effect of injury on sensory motor functions, memory and anxiety were assessed by beam walking, radial arms maze and elevated plus maze at 3–7 days following TBI. The only behavioral deficits found were transient impairments in working and reference memory. Blood serum was analyzed at 1, 3, and 14 days after injury for changes in selected protein biomarkers. Serum levels of neurofilament heavy chain and Tau, as well as S100B and myelin basic protein showed significant increases in the injured animals at all time points. No signs of macroscopic injuries such as intracerebral hematomas or contusions were found. Amyloid precursor protein immunostaining indicated axonal injuries at all time points analyzed. In summary, this model mimics some of the key symptoms of mTBI, such as transient memory impairment, which is paralleled by an increase in serum biomarkers. Our findings suggest that serum biomarkers may be used to detect mTBI. The model provides a suitable foundation for further investigation of the underlying pathology of mTBI.

Memory maintenance by PKMζ — an evolutionary perspective

Long-term memory is believed to be maintained by persistent modifications of synaptic transmission within the neural circuits that mediate behavior. Thus, long-term potentiation (LTP) is widely studied as a potential physiological basis for the persistent enhancement of synaptic strength that might sustain memory. Whereas the molecular mechanisms that initially induce LTP have been extensively characterized, the mechanisms that persistently maintain the potentiation have not. Recently, however, a candidate molecular mechanism linking the maintenance of LTP and the storage of long-term memory has been identified. The persistent activity of the autonomously active, atypical protein kinase C (aPKC) isoform, PKMζ, is both necessary and sufficient for maintaining LTP. Furthermore, blocking PKMζ activity by pharmacological or dominant negative inhibitors disrupts previously stored long-term memories in a variety of neural circuits, including spatial and trace memories in the hippocampus, aversive memories in the basolateral amygdala, appetitive memories in the nucleus accumbens, habit memory in the dorsal lateral striatum, and elementary associations, extinction, and skilled sensorimotor memories in the neocortex. During LTP and memory formation, PKMζ is synthesized de novo as a constitutively active kinase. This molecular mechanism for memory storage is evolutionarily conserved. PKMζ formation through new protein synthesis likely originated in early vertebrates ~500 million years ago during the Cambrian period. Other mechanisms for forming persistently active PKM from aPKC are found in invertebrates, and inhibiting this atypical PKM disrupts long-term memory in the invertebrate model systems Drosophila melanogaster and Aplysia californica. Conversely, overexpressing PKMζ enhances memory in flies and rodents. PKMζ persistently enhances synaptic strength by maintaining increased numbers of AMPA receptors at postsynaptic sites, a mechanism that might have evolved from the general function of aPKC in trafficking membrane proteins to the apical compartment of polarized cells. This mechanism of memory may have had adaptive advantages because it is both stable and reversible, as demonstrated by the downregulation of experience-dependent, long-term increases in PKMζ after extinction and reconsolidation blockade that attenuate learned behavior. Thus, PKMζ, the “working end” of LTP, is a component of an evolutionarily conserved molecular mechanism for the persistent, yet flexible storage of long-term memory.