Perhaps, it is an entrance to first specify how codes of communication operate before addressing the more difficult question of how symbolically generalized codes operate. Codes operate latently.
For example, when one submits a manuscript to a journal, the editor and the referees use criteria for accepting or rejecting the submission. When the paper is accepted or rejected a code of communication has operated. This code is not symbolically generalized since journal/specialty/discourse specific.
When one analyzes a set of rejections and acceptances, one can reconstruct this latent code. Because the code is selective a skewed probability distribution is generated. What produces the skewness in the distribution? The code.
When one submits a payment through the banking system, this payment is not rejected if the signature is proper, the credit sufficient, etc., because the code is now symbolically generalized. However, the payment is not the code. The code is that which makes the payments being done. Not the specific bank, but the banking system. (L. Leydesdorff in der Luhmannliste, 23.5.08)
Programs can be considered as production rules; they follow the arrow of time. Codes are validation rules: they feedback against the arrow of time. In other words, they provide what happened historically (that is, the variation) with meaning from the perspective of hindsight.
Fragen: