← all systems · see it in The System
Learning
Learning is the process of acquiring new understanding, knowledge, behavior, skills, values, attitudes, and preferences.
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | kind: associative learning in or before the Cambrian (proposed); person: habituation by 32 weeks of gestation |
| born precision | kind: period-level and contested; person: weeks |
| boundary | Inside: the change in what an animal does that experience brings about, and the steps that make it: registering a stimulus, relating it to other events (pairing, contingency, timing), the teaching signals that say a prediction was wrong, the synaptic and cellular change, the learned change kept, and its expression in behaviour. Outside: the senses that deliver the events (Human's channels), memory as the store learning writes into (its own walk), sensory adaptation, fatigue and injury (excluded by the definition of non-associative learning), innate behaviour (non-learning), and the education, teaching and culture that are carried on it. |
| nested level | A faculty of nervous systems (contested below them): built from synapses and their plasticity, teaching-signal circuits (midbrain dopamine neurons, inferior olive and climbing fibres, insect dopaminergic neurons) and the circuits that hold what is learned (striatum, cerebellum, hippocampus, amygdala, prefrontal cortex); its kept results are part of memory; education, culture and traditions run on it. |
| scale | organism |
Measures (25) published science
| earliest human habituation | 32 weeks of gestation |
| short-term vs long-term stage (Aplysia) | one trial: minutes; repeated spaced training: days to weeks time |
| early vs late LTP | E-LTP 1 to 3 h (no protein synthesis); L-LTP at least 24 h |
| synaptic terminals after long-term learning (Aplysia) | habituation: 35% fewer connections; sensitization: over two-fold more terminals relative count |
| central neurons in learning model animals | about 20,000 (Aplysia) to 300,000 (Drosophila) neurons |
| optimal interval for long-term sensitization (Aplysia) | 45 (not 15 or 60) min |
| refractory interval for cumulative LTP | 60 or 90 min cumulative; 10 or 40 min not |
| shortest effective spacing (honeybee olfactory learning) | 1 min |
| spacing for six-month verbal retention (people) | 7 days better than 3 days |
| single-error change (cerebellum) | about 5% on the next trial; about 100-fold larger than the cumulative change; decays in about 6 s % ; s |
| trials to asymptote (cerebellar learning) | hundreds to thousands trials |
| cerebellar cortex dependence after training | essential for the first several hours; not essential 24 h later |
| dopamine salience component | a few tens to hundreds ms |
| exposure for infant word segmentation | 2 minutes, 8-month-olds |
| transitional probabilities within vs between words | 1.0 vs 0.33 probability |
| social spread of a seeded technique (great tits) | average 75% (68%-83%) of each local population solved at least once in 20 days % of birds |
| population turnover across which a tradition persisted | 40% of birds present the previous year |
| Paramecium retention after spaced training (Gelber) | 3 h (spaced, not massed); later up to 12 h |
| genes with altered methylation after fear conditioning | 9.2 % of hippocampal genes (rats) |
| pea conditioning, original vs replication | Gagliano: p=0.0027 (n = 13 + 10); Markel: p=0.335 (n = 61 + 60) Fisher's exact p |
| pea seedlings with a conditioned response (original study) | 62% (F+L), 69% (F vs L) % of seedlings |
| habituation of passion-flower tendrils (Darwin) | 54 h of training |
| experimental evolution of learning (flies) | 30 generations |
| children and chimpanzees tested | 90 children (4-5 years), 69 chimpanzees individuals |
| amblyopia prevalence | two to five percent of the population % |
Open contentions (7) open contention
Where sources or readings disagree. Both stay on the map.
- C-2534 What is really learned in learned helplessness?
booked: The map keeps the competing readings; none is ruled.
the other reading: the original 1967 theory (same authors): animals learn that outcomes are independent of their responses; stressors themselves cause the harm and control blunts it (named by Maier and Seligman as the reverse reading their data now support) - C-2535 Did learning drive the Cambrian explosion?
booked: The map keeps the competing readings; none is ruled.
the other reading: learning arose in the Cambrian without driving it; not yet tested - C-2536 When did learning first appear?
booked: The map keeps the competing readings; none is ruled.
the other reading: rung 6: single cells learn (habituation in Stentor; Gelber's Pavlovian conditioning in Paramecium, argued convincing by Gershman et al. 2021); plants learn (habituation in Mimosa; associative learning in peas, Gagliano 2016, not replicated by Markel 2020); the dispute is partly over definitions of learning (Abramson and Chicas-Mosier 2016) - C-2546 Do memories just fade on their own?
booked: The map keeps the competing readings; none is ruled.
the other reading: bug: a biological limitation; feature: lets the network adapt to a dynamic world - C-2547 Can plants learn?
booked: The map keeps the competing readings; none is ruled.
the other reading: plants learn by association (Gagliano); no evidence for associative learning in peas (Markel); habituation in Mimosa is better supported than conditioning (Abramson) - C-2556 Can single cells learn?
booked: The map keeps the competing readings; none is ruled.
the other reading: Jensen; Katz and Deterline: changes in the culture (residual bacteria), not in the cell; Gershman et al.: the stirring result is contingency degradation, as in animals - C-2564 Does unlimited associative learning mark consciousness?
booked: The map keeps the competing readings; none is ruled.
the other reading: learning, even open-ended, can occur without experience (the authors' own AI objection, answered by restricting the marker to living systems); consciousness is uniquely human or tied to language (markers the authors call unambitious)
Partners and parts published science
Built from / builds (11)
- Education
- Synapse
- Cerebellum
- Prefrontal cortex
- Hippocampus (memory)
- Inferior olive
- long term potentiation
- Midbrain dopamine neurons
- Gene expression
- Learning against fixed (unlearned) behaviour across generations
- The classroom and online learning competing for learners' hours
Sits inside / holds (7)
- Play
- Classical conditioning
- Operant conditioning
- Nervous system
- Human (Homo sapiens)
- Memory
- Gene expression
Trades with (valence) (29)
- Phonology: how languages organise the meaningless units their words are built from (phonemes in spoken languages; handshape, location and movement in signed ones), together with the field of linguistics that studies those systems via sys:acoustic-wave · Learning gain
- Play via sys:sensory-environment · Learning gain
- Motivation via sys:sensory-environment · Learning gain
- Classical conditioning via sys:sensory-environment · Learning gain
- Operant conditioning via sys:sensory-environment · Learning gain
- Morphology (linguistics): how a language builds its words from smaller meaningful parts (roots, affixes, compounding, inflection and derivation), how languages differ in this, how speakers learn and use it, and the field that studies it via sys:acoustic-wave · Learning even
- Consumer via sys:memory · Learning even
- Synapse via sys:aqueous-solution · Learning gain
- Paramecium via sys:sensory-environment · Learning gain
- Cerebellum via sys:electrochemical-gradient · Learning gain
- Human (Homo sapiens) via sys:sensory-environment · Learning gain
- Prefrontal cortex via sys:electrochemical-gradient · Learning gain
- Natural selection via sys:genealogy · Learning mixed
- Plant (kingdom Plantae) via sys:sensory-environment · Learning mixed
- Hippocampus (memory) via sys:electrochemical-gradient · Learning gain
- Education via sys:acoustic-wave · Learning gain
- Chimpanzee (Pan troglodytes) via sys:sensory-environment · Learning gain
- Memory via sys:electrochemical-gradient · Learning gain
- Striatum via sys:aqueous-solution · Learning gain
- Machine learning via sys:mental-representation · Learning even
- Midbrain dopamine neurons via sys:aqueous-solution · Learning gain
- Fruit fly (Drosophila melanogaster) via sys:genealogy · Learning gain
- Culture via sys:sensory-environment · Learning gain
- Great tit (Parus major) via sys:sensory-environment · Learning gain
- Teacher via sys:acoustic-wave · Learning gain
- Dorsal raphe nucleus via sys:aqueous-solution · Learning mixed
- Human development (the lifespan) via sys:sensory-environment · Learning gain
- Gene expression via sys:sensory-environment · Learning gain
- Intelligence via sys:electrochemical-gradient · Learning gain
expressed by (13)
- Quality Improvement/Refinement
- Evolvability/Plasticity Engineering (Design for Future Variation)
- Sensing/Scanning
- Uncertainty Reduction/Learning
- Intrinsic Motivation / Curiosity-Driven Sampling
- Active Experimentation / Probing (Optimal Experimental Design)
- Trail-Marking / Externalized Memory (Stigmergy)
- Surprise / Change-Point Detection
- Social Learning / Public Information Use
- Quick Harm Avoidance
- Self-Monitoring
- Validation of Effectiveness
- Triage & Prioritized Allocation
floor filled by (23)
- Phonology: how languages organise the meaningless units their words are built from (phonemes in spoken languages; handshape, location and movement in signed ones), together with the field of linguistics that studies those systems
- Play
- Education
- Morphology (linguistics): how a language builds its words from smaller meaningful parts (roots, affixes, compounding, inflection and derivation), how languages differ in this, how speakers learn and use it, and the field that studies it
- Synapse
- Sensory neuron
- Cerebellum
- Human (Homo sapiens)
- Prefrontal cortex
- Human brain
- The world through the senses (sensory environment)
- Hippocampus (memory)
- Memory
- Inferior olive
- long term potentiation
- Striatum
- Midbrain dopamine neurons
- Culture
- Great tit (Parus major)
- Gene expression
- Mechanoreceptor channel
- Human development (the lifespan)
- Intelligence
represents (1)
Kinds (21)
- Associative learning
- Play
- Statistical learning
- Machine learning
- Enculturation
- Classical conditioning
- Operant conditioning
- Habituation
- Sensitization
- Non-associative learning
- Observational (social) learning
- Imprinting
- Motor (supervised, error-based) learning
- Unlimited associative learning (UAL)
- Limited associative learning
- Learned helplessness
- Episodic learning
- Rote learning
- Meaningful learning
- Formal learning
- Informal learning
Notes published science
- Learned helplessness was found backwards. Fifty years after discovering it, its own discoverers concluded from the brain circuits that 'helplessness was not learned in the original experiments': giving up is the brain's default reaction to prolonged bad events, and what animals (and people) actually learn is that bad events can be controlled, a lesson held in the prefrontal cortex that switches the default off (src:maier-seligman-2016).
- Two wild great tits were enough to start a tradition. Trained to open a puzzle box one way, they seeded a technique that about three-quarters of their local population picked up within 20 days, and the birds kept choosing the seeded way even after discovering an equally good alternative. A year later, with 60% of the birds new, the population went straight back to its own tradition (src:aplin-2015).
- Practice has a speed limit set by chemistry. In sea slugs two shocks build the longest-lasting memory when they come 45 minutes apart, not 15 or 60, because each one has to feed a slow biochemical process separately; in people the best gap grows with how long you need to remember, and for words kept six months a week between sessions beat three days (src:smolen-2016).
- A word you have just learned does not yet compete with the words you already know. It starts to (slowing your recognition of look-alikes such as 'cathedral' after learning 'cathedruke') only after a night's sleep (src:davis-gaskell-2009).
- Students in lectures feel they learned more than students in active classes, but they learned less: a smooth lecture feels like learning, and that feeling helps keep lecturing in most college science classrooms. (src:deslauriers-2019, src:freeman-2014)
- The oldest curriculum we can reconstruct already ran easy to hard: Old Babylonian scribal pupils at Nippur began by copying the three single wedges of cuneiform, then sign lists, then noun lists, then whole Sumerian sentences, the same 'easy samples to hard ones' order machine-learning curricula use today, and a sequence built to hamper a neural network hampers people too (src:wp:edubba, src:soviany-2021, src:strittmatter-2026).
- Learning to read breaks a habit our visual system has had for a very long time: treating a shape and its mirror image as the same thing. People who can't read keep it. People who learned only Tamil, whose script has no mirror-image letters, don't fully lose it. Only adults who also learned the Latin alphabet, with b and d, broke it completely (src:fernandes-2021).
- A toddler learning to walk takes about 2,368 steps, covers 701 m (7.7 American football fields) and falls 17 times every hour; walking is learned through thousands of short, varied bursts of practice, not smooth repetition (src:adolph-2012).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
learning: from an experience to a lasting change in behaviour (reader's view)
- L1: The experience: events in the world and the body, single or repeated
- L2: Registration: receptors and sensory neurons; the first place a response can change (receptor potentials shrink in habituating Stentor; transmitter release falls at the Aplysia sensory synapse)
- L3: Relation between events: stimulus with stimulus (classical), act with outcome (operant), across a time gap (trace), and contingency (control is present when the outcome's probability differs with and without the response)
- L4: Teaching signal: prediction error. Dopamine neurons code reward minus prediction; climbing fibres carry movement errors; a fully predicted outcome teaches nothing
- L5: Plasticity: transmitter release turned up or down, cAMP-PKA-CREB signalling, memory-suppressor genes, protein synthesis, growth or retraction of synapses, LTP, epigenetic marks
- L6: What is learned, kept: associations, habits, skills and knowledge, short-term (minutes) and long-term (days to weeks); traces moved to more durable sites (cerebellar cortex to cerebellar nuclei within a day); circuits shaped in critical periods
- L7: Changed behaviour and its open-ended use: acting differently in old and new situations; in animals with unlimited associative learning, learning about the world and itself in an open-ended way (proposed as a marker of experience, contested)
the boardroom (shared floors)
- L4: Teaching signals: prediction errors. Midbrain dopamine neurons code received minus predicted reward (Rescorla-Wagner, temporal-difference learning); the inferior olive's climbing fibres carry movement errors whose potency is regulated; a first, unselective salience component lasts tens to hundreds of milliseconds
- L5: Plasticity: transmitter release turned down (habituation) or up (sensitization); cAMP, PKA and CREB-1 switch short-term change to long-term growth of synapses, held back by memory suppressors (CREB-2, phosphatases); E-LTP 1-3 h without protein synthesis, L-LTP 24 h and more with it; epigenetic marks; in single cells, if they learn, a non-synaptic store
- L6: What is learned, kept: a short-term stage of minutes and a long-term stage of days to weeks; habituated synapses retract (35% fewer connections), sensitized ones more than double; cerebellar traces move from cortex to nuclei within a day; critical and sensitive periods decide when circuits can take input
- L7: Changed behaviour and open-ended learning: the animal acts differently; with unlimited associative learning (most vertebrates, coleoid cephalopods, some arthropods, as conjectured) it learns about the world and itself in an open-ended way, proposed as a transition marker for experiencing; in people learning may be conscious or not
way in: A stimulus that repeats (habituation and sensitization)
- L1: The same stimulus again and again (a touch, a stuffed owl in the cage); or a harmful stimulus elsewhere on the body
- L2: Receptors and sensory neurons: in Stentor the mechanoreceptor channels' voltage dependence shifts and receptor potentials shrink
- L3: The first synapse: less transmitter released in habituation, more in sensitization and dishabituation (Aplysia sensory-to-motor synapse)
- L4: Modulatory input: a harmful stimulus releases serotonin onto the sensory neurons and acts through cAMP; a strong novel stimulus restores a habituated response (dishabituation)
way in: Being taught and told (instruction, formal learning)
- L1: A teacher's words, a text, a demonstration set up by an experienced individual
- L2: Hearing and reading in a teacher-student setting with goals set by others
- L3: Meaning: knowledge learned by repetition alone (rote) or understood as it relates to other knowledge (meaningful)
way in: Errors in one's own movement (supervised motor learning)
- L1: A mismatch between what a movement did and what was wanted (the eye slips off a target, a reach misses)
- L2: Input streams of many modalities arrive by mossy fibres and are recoded in the granule cell layer
- L3: Inferior olive neurons fire; their climbing fibres carry the instructive signal to Purkinje cells, with a potency regulated by several factors
way in: Inner: stress and control over bad events
- L1: An aversive event (shock, loud noise) that can or cannot be escaped
- L2: Aversive stimulation activates serotonin neurons of the dorsal raphe nucleus, which sensitize for a few days
- L3: DETECT: the prelimbic cortex with the dorsomedial striatum registers whether the response controls the outcome
- L4: ACT: other prelimbic neurons project to the dorsal raphe and inhibit it, preventing passivity
way in: Inner: the trace reactivated (retrieval, rehearsal, spaced practice)
- L1: A memory trace formed by an earlier trial, still there when the next spaced trial comes
- L2: Reactivation by the next trial; with massed trials the trace is still active and is not retrieved
way in: Regularities in the input (statistical learning)
- L1: A continuous stream: syllables of speech, tones, shapes, with no pauses marking the units
- L2: The ear (or eye) takes in elements; which elements count (absolute or relative pitch) depends on age and experience
- L3: Transitional probabilities computed between elements: 1.0 within words, 0.33 between them; frequency alone does not explain it
way in: Stimuli that come in pairs (classical, Pavlovian conditioning)
- L1: Two events in the world: a neutral stimulus (a bell, a fan's wind) followed by one that evokes a reflex (food, light)
- L2: The senses take in both stimuli (hearing the bell, tasting the food)
- L3: Timing and contingency: the neutral stimulus comes to predict the other; learning can bridge a gap with no overlap (trace conditioning); decorrelating the two removes the response even with the same number of pairings
way in: The outcome of one's own act (operant, instrumental learning)
- L1: A consequence in the world after an act: a latch opens and food is reached; a reward, or a punisher
- L2: The senses register the outcome; only the learner's own act makes it come
- L3: Act-outcome relation: responses kept when they lead to a successful outcome, dropped when they do not; the schedule of reinforcement shapes the rate
way in: What other individuals do (observational and social learning)
- L1: Another animal's behaviour or its products: a demonstrator opening a puzzle box one way
- L2: The observer watches; in people, no reinforcement is needed, only a model
- L3: Copying rules: conformity to the local majority (great tits), weighing others' payoffs against one's own (children) or one's own only (chimpanzees)
way out: A changed reflex or bodily response (habituated, sensitized or conditioned)
- L1: The response in the world: salivation to a bell, a withdrawal, a turn of growth toward a fan (contested in plants)
- L2: The motor neurons get more or less input; the reflex is weaker or stronger
- L6: The kept change at the sensory-to-motor synapse (fewer or more connections)
way out: A chosen act, more or less likely (operant)
- L1: The act on the world: pressing the same button again
- L7: The act chosen again because it paid more than predicted, avoided because it paid less
way out: A recalibrated movement (skill)
- L1: The movement made again with a smaller error; skill grows over hundreds or thousands of trials
- L6: The adjusted mapping, kept first in cerebellar cortex and within a day in the cerebellar nuclei
way out: Active coping or passivity in the face of bad events
- L1: Escape attempted, or not: running in the shuttlebox
- L7: EXPECT: after control, the prelimbic-raphe pathway is changed over hours and later treats even inescapable stress as controllable
way out: Behaviour performed where others can learn it (traditions, teaching)
- L1: The act seen by others, the input of their 'others' channel
- L7: A learned technique used day after day; it spreads through the social network and becomes a local tradition
way out: Knowledge used and passed on in words (people)
- L1: Speech, writing and demonstration taken in by others: teaching, the demonstrator's side of social learning
- L7: Cognitive domain: to recall, calculate, discuss, analyse, solve problems
Sources (30) published science
Schultz W (2016) Dopamine reward prediction error coding. Dialogues Clin Neurosci 18:23-32. doi:10.31887/DCNS.2016.18.1/wschultz (PMC4826767, open) (behind its links)
Smolen P, Zhang Y, Byrne JH 2016. The right time to learn: mechanisms and optimization of spaced learning. Nat Rev Neurosci. PMC5126970 (behind its links)
Maier SF, Seligman MEP 2016. Learned helplessness at fifty: insights from neuroscience. Psychol Rev. doi:10.1037/rev0000033 (behind its links)
Dunlap AS, Stephens DW 2009. Components of change in the evolution of learning and unlearned preference. Proc R Soc B. doi:10.1098/rspb.2009.0602 (behind its links)
Kandel ER 2012. The molecular biology of memory: cAMP, PKA, CRE, CREB-1, CREB-2, and CPEB. Mol Brain. doi:10.1186/1756-6606-5-14 (behind its links)
Play (activity) (Wikipedia, rev 1381836613, 2026-10-25) (behind its links)
Classical conditioning (Wikipedia, rev 1380298957, 2026-10-14, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Operant conditioning (Wikipedia, via Wikimedia Enterprise snapshot) (behind its links)
Regulation of gene expression (Wikipedia) (behind its links)
Saffran JR, Kirkham NZ (2018) Infant statistical learning. Annu Rev Psychol 69:181. doi:10.1146/annurev-psych-122216-011805 (behind its links)
Gershman SJ, Balbi PE, Gallistel CR, Gunawardena J 2021. Reconsidering the evidence for learning in single cells. eLife. doi:10.7554/eLife.61907 (behind its links)
Markel K 2020. Lack of evidence for associative learning in pea plants. eLife. doi:10.7554/eLife.57614 (behind its links)
Gagliano M, Vyazovskiy VV, Borbely AA, Grimonprez M, Depczynski M 2016. Learning by association in plants. Sci Rep. doi:10.1038/srep38427 (behind its links)
Abramson CI, Chicas-Mosier AM 2016. Learning in plants: lessons from Mimosa pudica. Front Psychol. doi:10.3389/fpsyg.2016.00417 (behind its links)
Vale GL et al. 2017. Testing differential use of payoff-biased social learning strategies in children and chimpanzees. Proc R Soc B. doi:10.1098/rspb.2017.1751 (behind its links)
Hensch TK, Bilimoria PM 2012. Re-opening windows: manipulating critical periods for brain development. Cerebrum. PMC3574806 (behind its links)
Mery F, Kawecki TJ 2003. A fitness cost of learning ability in Drosophila melanogaster. Proc R Soc B. doi:10.1098/rspb.2003.2548 (behind its links)
Social learning in animals (Wikipedia, rev 1378620350, 2026-10-05) (behind its links)
Kuhl PK (2010) Brain mechanisms in early language acquisition. Neuron 67:713-727. doi:10.1016/j.neuron.2010.08.038 (author manuscript) (behind its links)
Consumer behaviour (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Chollet F (2019) On the measure of intelligence. arXiv:1911.01547 (behind its links)