Trial & error is basic to acquisition of new knowledge

With some exceptions in situation where organisms’ have evolved special learning processes such as cognitive maps or maternal imprinting, virtually all new knowledge comes from a variation on trial and error learning. Even if I can figure out on the first try how long it will take a ball to fall from the roof of Conant Hall to the ground, it is because Newton spent years working on the problems. I only had to take a few physics classes.

Thorndike studied trial & error learning

(In doing so, he formulated the “law of effect.” He also was interested in the extent of observational learning and teaching in non-humans.)

The structure of trial and error learning

(this is just common sense.)

a. Encounter a Specific problem

I’m thinking of a broad range of problems – a chicken getting out of its egg, Thorndike’s cats getting out of a box, Kohler’s chimp Sultana getting a banana, a child learning English, or a physicist solving a problem.

b. What's the behavioral repertoire – the set of possible responses as solutions?

I would define this broadly to include everything from Thorndike’s cats’ motor responses to a scientist’s store of hypotheses.
Some writer suggest trial and error be reserved for actual Sensorimotor actions but this neglects common structure and process in more abstract forms of gaining information by manipulation of representations in place of objects, or some combination of these.

c. Selecting a specific response to evaluate

This is a critical step – especially where the number of possible responses is large. There are several possibilities:

1. select at random

2. rely on “instinct” – perhaps species are provided by evolution with stimulus-response reflexes that lead to important adaptive solutions. For example, a chicken within an egg has a very limited movement repertoire – move its head a bit which jams its beak sufficiently – at the right time -- into the egg to crack it. This example suggests how “intelligence” is related to trial and error processes – little intelligence is required as the repertoire is so limited by evolution that the individual has no real options.
In other cases – in particular those involving social priming – the behavior of others focuses attention on elements of the problem solution, priming one response over the alternatives available.

3. rely on past experiences in similar situation. Psychologists since Thorndike have studied this in great detail – even the dullest organisms can learn to vary their responses according to the current stimulus environment.

4. Create a new response that extends the repertoire. Creativity usually comes down to combining existing responses into novel ones – but admittedly this is one of psychology’s big problems to solve.

5. Learn someone else’s responses to that problem and use them. Humans are good at this; that’s what schools are for (see Piaget on “Formal operations.) Even though I’m not anywhere as intelligent as Newton or Einstein, I can use the “mind tools” they created and solve certain problems perhaps as well as they could. (You don’t have to look to physics for examples here – everyday I use words and numbers that were created by unknown humans that enable much of what I do each day.)

d. have a method of evaluating a given response --success or fail?

This may be obvious in the case of a cat getting out of the box or more subtle in scientific experiments where statistics may be required.

e. have a memory record of outcomes that governs subsequent trials

Obviously the law of effect could not work with an organism that could not store new information in some way. Blocking new memories blocks new learning. At the other extreme, we know how important it is in scientific research to keep good records. As the young woman in the Jay Leno“Pentagon” feature demonstrated, even simple trial and error won’t work if you can’t keep track of previous trials!

Some exceptions to trial and error learning?

Cognitive maps (PC and Animal Intelligence video)

Maternal Attachment and imprinting
Many of you already know of Harry Harlow’s work on attachment, or Lorenz’s findings about “imprinting,”

Human language and other communication systems
In situations where apparently complex learning occurs rapidly and uniformly in all members of a species, it is reasonable to suspect that there really is less real learning than it appears. In the case of human language, our brain and vocal tract have evolved to enable very rapid language development without few (though notable) errors. See below.

What about socially mediated learning?

Social priming

Most cases of non-human socially mediated learning seem to be instances of social priming, followed by a restricted amount of trial and error. In the many primate examples seen in videos, mothers appear to draw their infant’s attention to objects, actions, and consequences, reducing the “trial space” needed to adaptively attain goals. Some see a connection between social learning, group size and brain size.


Imitation appears to be a fine-grained variation on social priming. In some cases, primate brains seem to be designed to enable one individual to mimic precisely the movements of a conspecific (one of its species).

Many human children around a year or so can mimic any speech sound they encounter without any notable trial and error.

Monkeys hand movements are primed by watching another manipulate a desirable object. See the “mirror neuron” video and notes.(PBS)


There are two points to consider here – first, anything that is taught had to be learned for the first time. So even if there is no trial & error in teaching, the information probably originated in trial & error.
Second, even in formal instruction, there is probably a certain amount of trial & error learning.


Final comment- many factors conspire in the "intelligence" of an individual. Diet, for example, may play a large complex role.