Below is a summary of the major ideas found in the papers by Spelke and Gallistel et al.

E. Spelke, "Initial Knowledge: Six Suggestions." Cognition 50(1994), 431-45.

< means `precedes'

Initial Knowledge Overall:

Domain-specific: constrains in one domain have limited access to and influence on constraints in another (e.g., early physical constraint of object identity and integrity does not transfer to early numerical knowledge, where sets may be constituted by unassociated elements)

Task-specific: constraints differ by task in which they are implemented (predictive reaching on the basis of innate spatial constraints is affected by whether children can act on the events and objects or just observe them)

Reliability < Ubiquity: reliable but not consistently observed information precedes consistently observed but unreliable information. Salience is thus a learning principle, not an innate metaconstraint.

Modularity < Interaction: domain-specific knowledge with limited access to other domain-specific knowledge precedes the interaction of domains and bootstrapping of aspects of one domain up from another. Some specifics: modular psychological and physical knowledge can bootstrap each other; modular number knowledge can bootstrap weight and density; modular geometric knowledge can bootstrap number.

Specific Constraints

Physical Knowledge:

cohesion (objects move as connected, bounded unities)
continuity (paths of motion are continuous)
contact (objects do not engage in action at a distance)
continuity < inertia/gravity
form/shape < color/texture/location

Number:

1 to 1 correspondence
succession

Social:

persons choose acts
persons can act at a distance

Geometry:

2 lines can intersect/occupy the same point
Euclidean geometry is sufficient

Constraints on Animal Learning

From: C.R. Gallistel, A. Brown, S. Carey, R. Gelman, F. Keil, "Lessons from Animal Learning for the Study of Cognitive Development." The Epigenesis of Mind: Essays on Biology and Cognition, S. Carey and R. Gelman (eds.) Hillsdale: Erlbaum, 1991, pp. 3-37.

Overall:

Domain-Specific Privileging: not limits on domain-general processes; positive biases (specificity of animal capabilities)

Sensitive and Critical Periods: learning in domains is best at a certain biological time and impossible after a certain biological time (Cf. birdsongs, owl vision, bird navigation, Universal Grammar)

Initial Knowledge Matures: some constraints are available only at certain maturational stages (e.g., some innate aspects of birdsong come into play only at later stages of development: Cf. Universal Grammar)

Underdetermination of Initial Knowledge: universal knowledge is schematic; experience fills in (e.g., cardinals' own vocalizations affect elaboration of the innate song-template: Cf. Universal Grammar)

Averaging: animals determine some representations by best average over instances (e,g, rats choose the consistently richest food source relative to all instances -- best food source may have less food but also fewer competing rats)

Model/~Model Continuum: species and subspecies are more or less dependent on exposure to a model to learn in a domain (e.g., some birds are tied to an external model for a song, some produce song irrespective of model, some have properties of each: Cf. Universal Grammar) Specific Positive Biases:

Shape < Salience: for many animals, the shape/layout of the environment organizes spatial learning, not salient features (Cf. Spelke for humans)

Rate, Proportion, and Magnitude: preference for food source can be constructed on the basis of these three, perhaps derivable from more basic computing of occurrences over an interval (Cf. Averaging)

Source of Information: species' and subspecies' learning can depend on nature of information source: e.g., some birds' song learning depends on acoustic properties of signal, others' on the kind of bird producing the signal (social source)

Dead-reckoning: many animals know their spatial position at all times (e.g., foraging ants, navigating birds)

Some Species-Specific Constraints:

Rats: pecking/digging/scrabbling links to food rearing and flight link to survival & shock NOTE: can't teach rats to rear at food

Taste: sweet < bitter, old < new, less < more

Confirmation from others: rats smell food on other rats' breath (Cf. ducks following others' choices of food source without sampling themselves)

Vervets: snake call/look to ground; leopard call/run into tree or stay there; eagle call/run into bushes look up. Schematic representation of: ground-threat, walk- threat, sky-threat?

Buntings: Seek and learn nighttime constellations and center of rotation of night sky (stars within 35% of stellar pole); individual variation in which constellations learned