IPL operates bidirectionally 

Integration of units of inner sensations are expected to provide a general framework of the memorized item (Vadakkan, 2010; 2013; 2019)

IPLs can be stabilized for different durations (Vadakkan, 2010; 2013)

Cue stimulus is provided access to all the unitary mechanisms (Vadakkan, 2009)

Dopamine may induce spine expansion (Yagishita et al., 2014) enhancing IPL formation. When IPLs last for long, it may lead to its stabilization. 

Glutamate induces spine enlargement in hippocampal slices (95%) (Matsuzaki et al., 2004) & in neocortex in vivo (22%) (Noguchi et al.,2019) - promotes IPL formation

IPL formation is an energy-intensive process, as demonstrated in experiments with artificial membranes (Rand & Parsegian, 1984; Martens & McMahon, 2008; Harrison, 2015).

Inter-LINKed spines within IILSPs can be shared by different stimuli to generate units of inner sensations that can be integrated to form memory (Vadakkan, 2010; 2013).

Repetition of learning, related learning, and learning events containing shared elements continue, coupled with the insertion of new neurons in the granule layer of the hippocampus, result in the formation of a surplus of new sparse IPLs in the cortex over time (Vadakkan, 2011). Eventually, net semblance for the item from the cortex alone will become sufficient to form memory of the item in response to a cue. 

Pre-existing inter-LINKed spines are utilized by common sensory elements present in a new learning event (Vadakkan, 2010a; 2013).

Inter-LINKing of spines that synapse to inhibitory inputs can regulate how specific inter-LINKed spines within an IILSPs can get selectively reactivated. Reversal of IPLs, deletion & formation of new spines also contribute. 

When a spine from each of two IILSPs gets inter-LINKed, every spine in the first islet forms a relationship with every spine in the second, thus increasing the search space. This allows generation of hypotheses about all possible relationships when a stimulus arrives to one of the inter-LINKed spines (Vadakkan, 2010; 2013).

A state of sleep is necessary to maintain postsynaptic depolarization induced by the presynaptic terminal as the system's dominant state. This dominance is crucial for enabling a cue stimulus to trick that postsynaptic terminal to hallucinate arrival of activity at its presynaptic terminal evoking memory (Vadakkan, 2016). See Minsky, 1980.  

Reduced sensory stimuli from environment is associated with lesser reactivation of inter-LINKed spines. This allows restoring the system to its baseline dominant state with lesser duration of sleep (Vadakkan, 2016).

Motor outputs can be voluntarily inhibited through inhibitory inputs to motor neurons while maintaining the ability of IPLs to generate inner sensations (Vadakkan, 2010; 2013)

It is not possible to voluntarily inhibit an IPL immediately. However, by a) inter-LINKing existing inter-LINKed spines with spines that receive inhibitory inputs, & b) rewiring the circuitry with feedback loops by new learning events (Vadakkan, 2007; 2010) it is possible to modify a memory.

Reactivation of IPLs during a higher brain function (which generates units of inner sensation) enables potentials to propagate from inter-LINKed spines to their postsynaptic neurons. If these potentials bring the neurons to threshold, they will fire (Vadakkan, 2010; 2016).

Exposure to the same stimulus long after learning will activate an additional set of neurons, through the propagation of potential across many IPLs formed from subsequent learning events. Also newly incorporated granule neurons introduce new paths that a cue stimulus propagates & fire neurons.

Learning & memory retrieval are associated with the activation of different sets of neurons.

Reactivation of IPLs formed between spines on overlapping dendrites of CA1 neurons provide additional potentials to their postsynaptic CA1 neurons that are being held at subthreshold activation levels. This explains firing of place cells (Vadakkan, 2013; 2016).

Changes in environmental stimuli, self-triggered thought processes, & various inner sensations such as fear, anticipation, hunger, & comfort fluctuate moment by moment, forming & reactivating new sets of IPLs. These continue to modify the network activity (Vadakkan, 2019).

Recent modeling studies have shown that a pyramidal neuron can fire an action potential through spatial summation (simultaneous summation) of nearly 140 EPSPs at the axonal hillock, originating from randomly located dendritic spines (Palmer et al., 2014; Eyal et al., 2018). However, based on energy calculations per bit of information, around 2,000 synaptic inputs are required for neuronal firing (Levy & Calvert, 2021). Both the degeneracy of inputs in firing a neuron & propagation of depolarization across the IPLs can explain the observed decorrelated firing events among neighboring cortical neurons. 

Contribution of potentials from each of the inter-LINKed spine (depending on nature of neighboring inter-LINKed spine's inputs) determine whether their subthreshold postsynaptic neurons fire or not. Limited number of muscles in the body must execute a vast array of motor outputs in response to a large number of sensory inputs. Hence, degeneracy of inputs in firing a neuron, and combinatorial motor unit activation enhance efficiency (for example, the muscles of the face & tongue generating speech). 

The subthreshold value determines the quantity of inputs that needs to reach those neurons to fire them. This enables control of motor outputs that for behavior and speech. 

Expectation of IPLs to form between the head regions of abutted spines that belong to different neurons (Vadakkan, 2010; 2016). 

Potentials generated within a large IILSPs of inter-LINKed spines that synapse with mainly with excitatory inputs explain voltage of a dendritic spike (Vadakkan, 2016).

Potentials from an IILSPs propagate to all the postsynaptic neurons of its inter-LINKed spines. The potentials reaching those neurons may not always become sufficient to cross the threshold for firing. 

The net potential of a dendritic spike may drain through some of the inter-LINKed spines to their respective neuronal cell bodies that are not being recorded (Vadakkan, 2016). This can occur especially when one of the inter-LINKs of the spine of the neuron under examination is with a spine that synapse to an inhibitory input. 

Routinely arriving stimuli may become neither essential nor detrimental to survival. Semblances induced by them merge together to form C-semblance bringing novel, beneficial & deleterious stimuli to attention. 

Seizures can be explained as rapid, chain-like formation of IPLs in the cortex (Vadakkan, 2016). This explains development of sensory & motor features from adjacent cortical areas.

The lateral spread of seizures via rapid IPL formation across the sensory cortices provides a mechanism for the internal perception of various sensations (Vadakkan, 2016).

IPL structure can progress to fusion between spines that belong to different neurons leads to mixing of their cytoplasm, which in turn lead to lateral spread of pathological changes such as spine loss & ultimately neuronal degeneration, as seen in ALS (Vadakkan, 2016). 

Excessive excitation can result in the pathological conversion of IPLs (that are normally limited to inter-spine membrane hemifusion) into inter-neuronal inter-spine membrane fusion (Vadakkan, 2016). This process can explain the observed dye spread between neurons.

Inter-neuronal inter-spine fusion can leads to mixing of cytoplasmic contents between neurons. Given that the expression profiles of even adjacent neurons of the same type can differ (Kamme et al., 2003; Cembrowski et al., 2016), cytoplasmic mixing is detrimental to both neurons. Hence, homeostatic mechanisms will trigger to cause loss of spines to protect neurons from further damage (Vadakkan, 2016).

Perineuronal net proteins surrounding the spine heads in the CA2 region (Carstens et al., 2016) can inhibit IPL formation between spines of different neurons, offering an explanation for the region's resistance to seizures (Vadakkan, 2016).

HSV fusion proteins has the potential to cause rapid formation of large number of non-specific IPLs & lead to seizures. Conversion of IPL hemifusion to fusion state, cause mixing cytoplasms of different neurons. Since expression profiles of even adjacent neurons of the same type are different (Kamme et al., 2003; Cembrowski et al., 2016), homeostatic mechanisms lead to loss of spines involved in fusion. If not successful, it can lead to neuronal death leading to cognitive defects (Vadakkan, 2016).

Anesthetic molecules are expected to generate a large number of non-specific IPLs, linking multiples of IILSPs. This enhances the horizontal component of oscillating potentials, significantly lowering the frequency of extracellular oscillations. Consequently, both specific inner sensations & motor actions are suppressed (Vadakkan, 2016). 

Seizure pathology involves the rapid formation of numerous non-specific IPLs, as well as IPL fusion between spines, which can result in spine loss & even neuronal degeneration (Vadakkan, 2016).

PDS is a large EPSP is generated through a postsynaptic mechanism (Johnson & Brown, 1981). Since distal dendrites typically generate EPSPs with amplitudes around 10 mV (Spruston, 2008), & the maximum voltage of a PDS can reach up to 50 mV, spatial summation of multiple EPSPs presents a plausible mechanism for the PDS (Vadakkan, 2016).

The loss of dendritic spines & neurons leads to a reduction in the number of specific IPLs required to generate the distinct units of inner sensation associated with a specific memory (Vadakkan, 2016).

Since units of perception (perceptons) from IPLs generated in a temporal pattern from consecutive flickers overlap it results in a continuous percept (Vadakkan, 2015).

Stimuli from inside the border form a percept with a different depth than stimuli from outside the border. thereby defining the border. This explains border contrasting with the background (Vadakkan, 2015).

Stimulation of sensory pathways anywhere along the input path (e.g. retina), before their convergence in the visual cortex, can form IPLs generating perceptons (Vadakkan, 2015).

It is possible to discriminate between two odorants when sniffed at a 60-millisecond interval (Wu et al., 2024).

Perception occurs through the rapid, reversible formation of IPLs. This allows for the formation of new sets of IPLs upon the arrival of the second odorant, generating distinct perceptons for the perception of the second odor (Vadakkan, 2015).

Perception occurs by the rapid formation of IPLs (to form perceptons) between abutted spines in V1. Lack of consistency in the set of firing neurons in response to the same visual stimulus don't affect perception (Vadakkan, 2015). Firing of postsynaptic neurons depends on: a) additional depolarization reaching these neurons via the newly formed IPLs & b) the sub-threshold activation state of the neurons. 

Total reactive perceptual delay for a novel, unpredictable event = T_{motion-onset latency} = ≈ 80–120 ms). This matches with the delay of at least 100ms between retinal photoreceptor cell stimulation & conscious perception (De Valois & De Valois, 1991; Nijhawan 2008). By the time the perceptons of the first percept of the moving object are integrated (after T_motion-onset), the physical object has already moved a distance Δx = v · T_motion-onset. This provides a plausible explanation for the Fröhlich effect (Vadakkan, 2022).

During continuous rotation, a stable semblance of motion is maintained by fast IPL reactivation cycles (T_update). When the pattern is suddenly replaced, 2 parallel processes begin. 1) Decay of motion semblance: The active IPL network supporting the rotation representation decays with a time constant Δt_rev (~30–300 ms). 2) Formation of novel static semblance: Perception of random texture requires slow de novo IPL formation, with a latency on the order of T_{percept(novel)} ≈ 70–100 ms. This creates a temporal conflict within the postdictive integration window. For a period after the switch, brain has simultaneous access to 1) a decaying neural representation signaling continued rotation, & 2) an incomplete, new representation of a static pattern. The perceptual system resolves this conflict by assigning the definitive sensory evidence to the perceptual "now." The lingering rotational activity is interpreted as having occurred just prior to this "now," creating the vivid illusion of a backward jump.

FLE is nullified at a direction change because the flash acts as a spatial anchor that collapses the forward integration horizon (τ_proj ≈ 116–162 ms) of the moving object. At the moment of reversal, the old motion vector begins to deactivate within one fast update cycle (T_update ≈ 36–42 ms before the turn), while the new direction (a novel event) requires slow de novo IPL formation (T_motion-onset ≈ 80–120 ms). The flash’s slowly assembling perceptons (T_{percept(flash)} ≈ 70–100 ms) provide a static spatial tag at the turn vertex. When reversal occurs early (~26 ms after the flash), the forward integration is still labile, and no new directional signal has consolidated; the flash’s anchor therefore collapses the integration to the vertex, aligning the percepts of the flash & moving object. For later reversals (> ~67 ms), forward IPLs stabilize, a new direction emerges, & the latency difference is re-expressed, restoring the FLE. Thus, cancellation at a turn reflects dynamic re anchoring within a postdictive integration window.

Anesthetic agents are known to exert diverse actions, including functioning as GABA-A receptor agonists, alpha-adrenergic receptor agonists, NMDA receptor antagonists, dopamine receptor antagonists, & opioid receptor agonists (Kopp et al., 2009).

General anesthesia induced by anesthetic agents can be reversed by applying external pressure to an animal enclosed in a sealed container – achieved by increasing air pressure for terrestrial animals or water pressure for aquatic animals (Lever et al., 1971; Halsey et al., 1986).

Only reduced amounts of anesthetic agents are required to induce anesthesia in the presence of levodopa (Segal et al., 1990). 

There have been several reports of cognitive decline following surgeries involving the use of general anesthetic agents (Baranov et al., 2009).

The rapid chain formation of numerous non-specific IPLs, triggered by alterations in extracellular matrix (ECM) properties (e.g., markedly low serum sodium) or by factors that enhance neuronal excitability, leads to seizure generation (Vadakkan, 2016). This alters conformation of net semblance during background state, disrupting C-semblance (Vadakkan, 2010).

Associative learning involves multiple shared components. Also, new neurons are continuously integrated into the granule layer of the hippocampus, generating more IPLs in the cortex. Consequently, more learning experiences can lead to the formation of excess IPLs in the cortex (Vadakkan, 2013; 2019) that allows individuals with advanced education to tolerate more loss of IPLs before exhibiting cognitive impairments.

Units of first person property are generated at locations inter-LINKed spines, which are formed mainly in regions of convergence of inputs. Different sensations (sensory cortices), associative learning (hippocampus) &  emotions (multi-convergent areas) mainly occur at corresponding regions.

They clear spill over neurotransmitter molecules, recycle them for reuse by the neurons. Since nearly all spines have nearly 50% area not covered by astrocytic pedicels, each spine has the potential to form IPLs.

When fast or first arriving features of predator or prey reactivated IPLs & generated first-person inner sensations of the remaining properties, it conferred survival advantages. It was naturally selected & conserved throughout evolution (Vadakkan, 2020).

Inter-neuronal fusion may have induced adaptive responses to prevent progression of IPL structure to full fusion, which is necessary for the continued IPL function. Consequently, initial inter-neuronal pore formation is thought to have triggered an adaptive mechanism that restricts further fusion, playing a key role in the evolutionary success of organisms with this capability.

Can be explained in terms of propagation of depolarization across the IPLs between spines belonging to different neurons (Vadakkan, 2013). This also accounts for why only sparsely distributed neurons fire in a time-correlated manner.

The distinct mRNA profiles of adjacent neurons, even of the same type, suggest that any mixing of cytoplasmic contents would trigger homeostatic mechanisms, such as spine loss, to prevent further damage (Vadakkan, 2016). This aligns with the structural limitation of IPLs to hemifusion.

Pathological fusion is responsible for neurodegeneration. Clinical features are determined by the formation of non-specific IPLs at different locations, & the locations of IPL fusion, which can lead to spine loss & even neuronal death (Vadakkan, 2016). This explains the observed heterogeneity.

The formation of extracellular plaques can reduce the number of specific IPLs. Individuals with a surplus of specific IPLs can afford to lose a fraction of IPLs. However, individuals with only a borderline number of IPLs (just enough to generate specific memories) will be vulnerable to the effects of amyloid plaque accumulation in the ECM.

Formation of new unrelated learning increases input combinations to a neuron. Enlargement of IILSPs, inter-LINKs with spines that synapse with inhibitory inputs, insertion of new granule neurons altering the circuitry can lead to the formation and reactivation of new set of IPLs. As a result, when a brain function is repeated, it activates a new set of neurons (Vadakkan, 2019).

Golgi used a single oxidizing agent to pre-treat brain tissue prior to staining, while Cajal introduced an additional oxidizing agent during the same step. This suggests that a higher oxidation state limits the spread of the Golgi staining reaction across neurons, likely by blocking certain inter-neuronal channels. Notably, blood oxygenation level-dependent (BOLD) signals have been observed to peak in specific brain regions approximately 3 to 4 seconds after learning (Monti et al., 2010; Murayama et al., 2010), & most working memories tend to fade over time. These observations suggest that oxygen likely play a role in reversing learning-induced channels. These match with the properties of IPLs (Vadakkan, 2022).

With insertion of granule neurons, repeated instances of the same associative learning lead to formation of new sparse IPLs at higher neuronal levels (Vadakkan, 2011).

Fear conditioning is associated with enlarged synapses on the dendritic spines of LA neurons (Ostroff et al., 2010; Choi et al., 2021). 

Synapses on the dendritic spines of LA neurons exhibit a higher ratio of postsynaptic density (PSD) area relative to that of presynaptic structures (Ostroff et al., 2012).

Astrocytic pedicels cover nearly less than 50% of the perisynaptic area in approximately 60% of synapses within the CA1 region of the hippocampus (Ventura & Harris, 1999). Synapses devoid of astrocytic coverage emerge in the amygdala during the consolidation of Pavlovian threat conditioning (Ostroff et al., 2014).

A disconnect between dendritic depolarization & neuronal firing has been observed during fear conditioning (d’Aquin et al., 2022).

When mice were injected with a histone acetyltransferase (HAT) enzyme to enhance transcription, the strength of their fear memory increased (Santoni et al., 2024). The study also found that a) neurons in which HAT is overexpressed are the neurons that fire during memory retrieval, & b) optogenetic silencing of these specific set of neurons prevents fear memory recall.

HAT removes histones from DNA increasing transcription of proteins necessary to synthesize phospholipids that promotes exocytosis & facilitates IPL formation of its spines. Silencing the neurons will prevent motor output functions necessary for fear expression.

Firing of LA neurons becomes more synchronized through modulation of theta frequency within the LA (Pare´ & Collins, 2000). Synchronous oscillations in the theta & gamma bands are observed between the basolateral amygdala (BLA) & interconnected brain regions during the retrieval & consolidation of fear memories (Bauer et al.,2004; Seidenbecher et al., 2003).

Reactivation of IPLs within the IILSPs on the dendrites of LA neurons contributes vector components to the oscillations, which in turn integrate units of semblances for fear. This integration of both vector components and units of inner sensations are expected to be a system-wide process involving different brain regions.

Memory retrieval induces synchronized rhythmic activity between basolateral amygdala (BLA) & interconnected brain structures, accompanied by the reactivation of certain sets of neurons that are called fear engram neurons (Bocchio et al., 2017).

The subjective aspect of mind (qualia, experience) is explained as the integral of semblances generated by IPL reactivations. At the IPL reactivation stage, physical membrane events generate internal virtual sensory states that the system "feels. 

Several structural features provide favorable conditions 1) Synapse-dense cortical areas with tightly packed neuropil have adjacent spines (Kasthuri et al., 2015; Zhu et al., 2021; Gemin et al., 2021). 2) The dendritic arbors of these neurons display significant territorial overlap (Mizuseki et al., 2011; Bezaire & Soltesz, 2019; Iascone et al., 2020). 3) The sister branches on a neuron’s dendritic tree often avoid overlapping (Grueber & Sagasti, 2010). 4) On a dendritic branch, the mean inter-spine distance exceeds the mean spine diameter (Konur et al., 2003). 5) Over half of the spine surface area lacks ensheathment by astrocytic processes (Ventura & Harris, 1999). These observations favor the presence of abutted spines either in pairs, chains or clusters. They are either simply abutted (waiting for stimuli to arrive from the right associative learning event) or are already functionally inter-LINKed.

The coexistence of redundancy & increased population information requires a shared hub upstream of the neurons that can both show redundancy in neuronal firing as well as specificity in information storage. For redundancy to grow within hundreds of milliseconds during a trial, it is necessary to demonstrate progressive recruitment of something upstream of the neurons. 

Based on the semblance hypothesis, each new learning event leads to the formation of new IPLs that provide horizontal cortical connectivity. Each neuron receives direct synaptic inputs as well as inputs from lateral spread via IPLs within the IILSPs. IPL networks a) explain how learning can increase shared neuronal information without reducing total information & b) enable a single neuron to reflect network-level information. When recordings are obtained from one neuron, we are indirectly sampling the state of a larger IPL network. This supports the finding in the paper that redundancy increases with learning, since each neuron becomes more representative of the population state. In other words, within IPL networks, redundancy represents shared signal structure. Structured redundancy provided through IILSPs improves robustness and decoding reliability. Progressive recruitment of IPL pathways explains how redundancy can grow within hundreds of milliseconds during a single trial. In summary, the learning-induced increase in shared neural information without loss of coding capacity is consistent with what would be expected if large networks of IPLs form between dendritic spines and distribute activation across neuronal populations.

Significant territorial overlap between dendritic arbours of neurons (Mizuseki et al., 2011; Bezaire & Soltesz, 2019; Iascone et al., 2020), lack of overlapping between sister branches of a neuron's dendritic tree (Grueber & Sagasti, 2010), & mean inter-spine distance exceeds the mean spine diameter (Konur et al., 2003) favor IPL formation between abutted spines that belong to different neurons. Hence, during retrieval, the signals from CS propagate through IPLs & fire a subset of neurons that fire in response to freezing in fear.

The high energy delivered during LTP stimulation protocols induces dendritic spine expansion & promotes the formation of numerous non-specific IPLs. These IPLs establish multiple pathways through which synaptic potentials can reach the recording electrode, resulting in a potentiated response. Prolonged persistence of these IPLs explains sustained, long-term nature of LTP (Vadakkan, 2019).

High energy applied during LTP stimulation is expected to induce inter-neuronal & inter-spine hemifusion. Membrane fusion blockers inhibit stages of fusion, preventing the induction of LTP (Vadakkan, 2019). Also, they can block the fusion of vesicles (that transport AMPA receptor subunits to the lateral spine region) with the spine membrane decreasing number of IPLs formed.

Higher the number of converging inputs, more the density of abutting dendritic spines from different input pathways. This facilitates the formation of a proportionally greater number of non-specific IPLs, which are implicated in the induction of LTP (Vadakkan, 2019). Hippocampus has high density of inputs from all sensations. 

Formation of non-specific IPLs is expected in response to high-energy stimuli during LTP, kindling, & seizures. Pathological conditions that lead to membrane instability, increased neuronal excitability, & ionic imbalances can further amplify these changes (Vadakkan, 2019).

The SNARE protein has the ability to bring together repulsive membranes & overcome the energy barriers associated with curvature deformations during hemifusion (Martens & McMahon, 2008; Olkers et al., 2016). It generates the force necessary to pull abutted membranes together as tightly as possible (Hernandez et al., 2012). The t-SNARE protein syntaxin plays a crucial role in generating local membrane trafficking within spines & directing membrane fusion (Kennedy et al., 2010).

AC: Auditory cortex

AMPAR: Fast responding glutamate receptors

ECM: Extracellular matrix (matrix between the brain cells)

EPSP: Excitatory postsynaptic potential

FLE: Flash-lag effect

HAT: Histone acetyltransferase

IILSPs: Islets of inter-LINKed spines  

IPL: Inter-postsynaptic functional LINKs (Inter-spine LINKs). These do not occur between adjacent spines on one dendritic branch

LA: Lateral amygdala

LINKs: links (Written in capital letters to show their significance)

LTD: Long-term depression

LTP: Long-term potentiation

MGN: Medial geniculate nucleus

MSN: Medium spiny neuron

NAc: Nucleus accumbens

PDS: Paroxysmal depolarization shift

STP: Short-term potentiation