Why is connectionism described as parallel-distributed processing?

Connectionism is described as parallel-distributed processing because the term highlights two fundamental features of how these computational models function:1

• Parallel: Unlike rule-based models that process information in a step-by-step, sequential fashion, connectionist systems operate through a large number of simple processing units—modeled after neurons—that work simultaneously. Computational tasks are performed collectively by generating firing activity across the network rather than through a series of individual inferences.21
• Distributed: Information is not stored in one specific location within the network. Instead, it is stored in a distributed fashion, where each individual connection participates in the storage of many different items of information. The knowledge required to perform a task is spread throughout the entire pattern of connection weights forged during training.1

By utilizing these mechanisms, connectionist models simulate how the human brain processes information, allowing for complex tasks like facial recognition and language learning to be accomplished through the simultaneous satisfaction of multiple constraints.21