A Sea of Paradigms

Note that the paradigms in this list are not necessarily mutually exclusive. Also, most languages or paradigms mix multiple concepts.


This is viewing organizations and/or code in a highly hierarchical manner. Most paradigms use at least some hierarchies, but the concept can be taken much further. Data, algorithms, and/or their relationship can be hierarchical.

For example, there are special database systems to process and control hierarchical data more effectively than RDBMS, at least from a performance standpoint. Pascal is known to provide a very hierarchical algorithm and subroutine organization. Object Oriented Programming (below) tends to use inheritance hierarchies to classify "objects". I find hierarchical taxonomies very limited for business applications even though superficially it seems like it would be more applicable.


This is my favorite. It can focus on either making programming for tables and RDBMS easier, and/or on using tables themselves to organize program logic.

Another feature of Table Oriented Programming is achieving a consistent set of collection processing operations.


This is focusing on grouping organizations and/or code by similar operations or by the operations that can be done on them.


This is organizing by objects or similar categories (subclasses) and a predetermined set of operations that operate on them. It is my opinion that Object Oriented Programming is over-hyped at this time. It has it's place, but not at the center of the Paradigm Universe.

Expert Systems

This organizes behavior by a list or pool of rules that are sometimes constantly cycled through. The rules can have priorities and may be "switched-off" under some circumstances and conditions.


These are popular in UNIX utilities. Stream paradigms provide solutions by linking the input and output of smaller programs or utilities, sort of like a bunch sausages linked together. Although streaming is highly useful for batch operations, they are more limited in highly interactive systems.

It is also tradition that the streams be text-based if possible. That makes them both easier to inspect and cross-platform.

See Also: Control Panel Analogy


This emphasizes starting with or focusing on the user interface and flushing out the rest from there. It may also emphasize logic building by dragging, dropping, pointing, pasting, etc.


This is using set theory, such as union, intersection, etc. to solve some types of problems. Set theory can be used in SQL, for example, to solve complex processing problems.


Lotus Notes is an example of a system that uses documents as the base for work-flow and other business tasks.

Circuit-Like Visual Logic

Lab View is a product that uses a circuit board-like layout to create program logic. Although the product was originally designed for electronic applications, programmers found that it could be used for many different application types. Subroutines are often represented by a chip-like rectangle and parameters are fed to these subroutines by "wiring" output from one "chip" to the input terminals of another. Special constraints can be set up to control the sequence and mode of the "chips".

Users familiar with it often prefer it over traditional languages because it provides a two-dimensional way to layout logic instead of the one-dimensional approach inherent in text coding. (Control Tables also offer a 2D view of logic, but in a different way.)

Neural Networks

The human mind uses neural networks. Although most practical applications tend to use neural networks for pattern matching and finding, it is possible they could be used for much more.

Neural networks are based on the concept of nodes with connections (wires) coming in and going out to many other nodes. The in-coming wires often have little "gates" on them that have a weighting factor. The weighting factor controls how much of the signal from a given wire gets "heard" from it's corresponding node.

The weighting factors are usually modeled with a range of either 0 to 1, or from -1 to 1, and is the primary way that neural networks are "programmed". (Behavior within a node can also be changed, but this is less common.)

Functional Programming

Not to be confused with "procedural programming", functional programming is where side-effects of functions on the state of the rest the program are impossible, or discouraged (depending the purity of your faith). Since side-effects-free functions don't have any context, it's easier to test them interactively (typical for Lisp systems), and also easier for compilers to optimize precisely because of the lack of side-effects. Lastly, 'refactoring' in functional languages is allegedly easier, again due to the lack of context. (Text based on notes from Philip Lijnzaad.)


Please let us know if we forgot any.

Updated: 3/22/2000, © Copyright 1999 by Findy Services and B. Jacobs