4.9 Signal Selection

Signal selection is the process by which a sender chooses a particular signal from among a repertoire of possible signals to transmit a message. In the framework of cybernetic communication theory, this selection process is the fundamental act that connects the informational content of a message to its physical representation. Shannon's original model treats the transmitted signal as the output of a selection made by a source: the source selects one of many possible messages, and the encoding maps that selection into a signal. The information carried by the signal is a function of the range of possible selections and their respective probabilities.

Shannon explicitly framed communication in terms of selection from a set. He described the operation of communication as the choice of a message from a set of possible messages, and the communication system's task as reproducing that choice at the destination. The informational content of a signal is therefore not intrinsic to the signal itself but depends on the statistical context of the selection: what other signals could have been selected, and with what probabilities. A signal that is always selected carries no information, while a signal that is rarely selected but has been chosen carries a great deal.

The number of possible signals in a repertoire, and their statistical structure, determine the capacity of a source to carry information. For a source that selects among N equiprobable signals, the information per selection is:

measured in bits. This grows logarithmically with the number of alternatives, reflecting that doubling the repertoire adds exactly one bit of potential information. When the signals are not equiprobable, the average information per selection is reduced from this maximum and equals the Shannon entropy of the probability distribution over the signal set:

Signal selection is not a purely mechanical or arbitrary process. In natural communication systems, signal selection is structured by a code, which maps meaningful states or intentions to specific signals from the repertoire. This code creates a correspondence between the space of meanings and the space of signals. The selection process then operates in the meaning space, with the code translating each selection into the corresponding signal for transmission. Senders and receivers must share compatible codes for communication to succeed.

In cybernetic systems, signal selection occurs at multiple levels simultaneously. At the lowest level, physical signals are selected from continuous or discrete alphabets of waveforms. At higher levels, symbols, words, or commands are selected from semantic vocabularies. Each level of selection contributes to the overall information-carrying capacity of the communication act. Redundancy between levels, where selections at a higher level constrain possible selections at lower levels, reduces the effective freedom of the system and hence the information per symbol but can improve robustness against noise.

The statistical properties of signal selection have direct consequences for the efficiency of communication. If a sender consistently selects certain signals more frequently than others, a well-designed code can assign shorter representations to common signals and longer ones to rare signals, achieving compression that approaches the entropy bound. Huffman codes and arithmetic codes implement this principle optimally, ensuring that the average code length per selected signal approaches the entropy of the source distribution.

Shannon's insight that the meaning of a message is irrelevant to its information content rests entirely on this selection-based framework. Two signals convey the same amount of information if they are selected from equivalent statistical contexts, regardless of what they mean or how they are interpreted. This separation of information from meaning is what makes information theory applicable across all communication systems, from electronic circuits to linguistic exchanges, and is the foundation upon which channel capacity, source coding, and error correction are all built.

The dynamics of signal selection in social communication introduce additional complexity not captured by the purely statistical model. Human communicators select signals based on pragmatic, contextual, and strategic considerations, not just statistical frequencies. The selection of one word over another, one gesture over another, or one framing over another in a human conversation reflects goals, relationships, and interpretive expectations that cannot be reduced to probability distributions. Cybernetic communication theory acknowledges this complexity by recognizing that the statistical model of signal selection captures only the syntactic and probabilistic dimensions of communication, leaving semantic and pragmatic dimensions to be analyzed with additional frameworks.