Levels of analysis

From RationalWiki
Jump to navigation Jump to search
Poetry of reality
Icon science.svg
We must know.
We will know.
A view from the
shoulders of giants.

Levels of analysis[note 1] refers to a way of analysing or designing a complex system. Here, a complex systemWikipedia[1] can be anything from an organism to a crystal, from a supernova to a culture, from a whirlwind to the human visual cortices. In specific cases, emergent properties are observed when moving upwards from a lower to a higher level of analysing a complex system. Starting from a sufficiently simple layer, each new layer adds complexity, organisation and (in many cases) an emergent phenomenon.[2]

Abstraction layers are deployed in cognitive science, in computer science, and in the natural sciences. In computer systems engineering, they usually prefer calling an individual level a "layer of abstraction".Wikipedia On the other hand, in the social sciences the term is usually "level of analysis".Wikipedia Regardless, the term refers to the method of discussing a complex dynamical system, in which meaningful hierarchical distinctions can be made. What that means is that understanding something on the cell-level, for example, mitosis, does not (on its own) help one understand things on the human-level and vice-versa. So, clearly, there must be some difference between the two levels of analysing what is essentially the same system: a human being. This is why the general concept of levels of analysis is useful, because it allows one to dissociate (given evidence) the various levels. In essence, it is a type of general purpose super-theory. It also allows, in specific cases, to demarcate causality from one level to another, e.g., a single cell dying in a multicellular organism is unlikely to have any causal effect on what the organism wants to have for tea.


For example, a desktop computer can be subdivided into the following simplified layers of abstraction: the electronics, logical gates, adders and (de)multiplexers; the hardware, HDDs, the CPU and RAM; and the software, word-processors, web-browsers and music-players. When discussing web-browsers there is no need to care about logical gates or what part of RAM is in use, because it is impossible to reduce higher layers to lower ones although anything that occurs at one layer reverberates to all others. The ability to abstract away, ignoring, the components of the implementation of lower layers, provided the exchange of information that occurs where layers meet conforms to specified standards, is known as the principle of multiple realisability (MR). Computers can be made of vacuum tubes, neurons, DNA or quantum computing chips and are proven to be computationally equivalent.[3]

Another famous example is David Marr's levels[4] used in neuro-, cognitive-, and computer science. Marr originally developed this organisational structure for both the understanding of, and the creation of models for, the human vision system. Although many other cognitive scientists developed their own levels of analysis for the same or similar purpose,[5][6][7][8] Marr's version remains the most wellknown and widely used.

Knowing about levels of abstraction also allows one to spot serious problems with (quasi)scientific reasoning. Good examples of such problems of reasoning can be found in evolutionary psychology, where the interactions of genes are proposed to interfere significantly with the workings of the brain and mind. Something which is a classical mistake in understanding abstraction. Once a brain is created, so given the genes produce an average brain (no genetic mutations that are harmful to development), it is to a great extent free from lower levels and is largely shaped by culture, society, and personal experience.[9]

See also[edit]

External links[edit]


  1. Also called layers/levels of abstraction, of description, or integrative layers/levels — so any sensible permutation.


  1. Complex systems on Scholarpedia.
  2. Theory of integrative levels. The British Journal For The Philosophy Of Science, 5(17), 59-66.
  3. Turing, A. (1937). On computable numbers, with an application to the entscheidungsproblem. Proceedings Of The London Mathematical Society, 42, 230-265.
  4. Marr, D. (1982). Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. New York, NY, USA: Henry Holt and Co., Inc.
  5. Fodor, J. A. (1975). The language of thought.
  6. Pylyshyn, Z. W. (1989). Computation and Cognition: Toward a Foundation for Cognitive Science.
  7. Anderson, J. (1990). The adaptive character of thought. Lawrence Erlbaum.
  8. Newell, A. (1990). Unified theories of cognition.
  9. Human nature, a Humean take by Massimo Pigliucci