Building Mental Representations for Software Engineers: From Architecture to Patterns

Drawing on the concept of deliberate practice from Peak: Secrets from the New Science of Expertise , the article argues that expert software engineers build effective mental representations—pre‑existing information patterns such as facts, images, rules, and relationships—rather than relying on passive learning or memorization.

Core Types of Mental Representations in Software

Architecture (Arch) : Visual, visualizable thinking product showing a software system's basic elements and relationships, including module identification and API identification.

Concept : Textual and visual abstraction derived from capturing nouns and commonalities.

Time Flow : Visual timeline‑based representation answering "How to do?"

Spatial Structure : Visual, inductive representation of spatial relationships—parent/child, sibling, peer, collaborative.

Decision Knowledge : Textual and visual deductive output expressed via mind maps, decision trees, or written records.

Historical Evolution : Textual and visual navigation through time, tracking how the system changed.

Concern : Textual and visual multi‑perspective, divergent thinking covering cross‑cutting aspects such as security, performance, reliability, observability; relates to tangling and scattering in software.

Feature/Function : Textual or visual inductive abstraction of core software capabilities, crucial for understanding a codebase.

Pattern : Textual and visual abstraction encompassing architectural styles, design patterns, idioms, algorithms, data structures, and rules for consistency, reliability, performance, etc.

These representations are formed through abstract thinking, visualization, inductive reasoning, deductive reasoning, and spatiotemporal thinking. The resulting text, visual images, and relationships are stored in long‑term memory; repeated updating of this neural network constitutes deliberate practice.