Thanks, Amos. As for you not knowing 'counfounding', sorry, my mistake. I have misspelled the word 'confounding', but since even with the correct spelling the word may be unknown here's what it is:

Undesired factors that additionally differentiate the experimental and control conditions are called confounding factors

It is an often overlooked problem that arises if you vary one factor and another factor (you even may not have thought of) is covaried as well. Then the result could be due to the factor you are interested in or to something very different. The validity of your conclusions could be in jeopardy.

An example from another field: There is a small but significant excess of incidence of schizophrenia among those born in winter (6 winter months compared to 6 summer months). Nobody believes that winter/summer itself is causal or that the sun signs of winter cause more schizophrenia. So the winter/summer difference must be confounded with another (not yet identified) variable: differences in food intake (varying with time of the year) of mothers during pregnancy, differences in first experiences in cold/warm times (the first steps of summer borns will be in summer which is a much better time for learning to walk), differential birth rates of different social backgrounds (in Germany, e.g., the nine-month-after carnival baby boom nearly exclusively happens among the unskilled workers)....

A true experiment would unconfound these factors by random assignement of participants to the conditions. So one half of the participants would have to give birth in winter and the other in summer. Describing that scenario suffices to make clear why a true experiment on this field should and will never be run. In many areas using real life data, true experiments cannot be done and, therefore, confounding is a big problem for real life data gathering.

Wolfgang