The disease model states that problems are symptoms of a small set of underlying disorders. This explains observable clinical symptoms by a small set of latent variables (e.g. depression). A network is a set of elements (nodes) connected through a set of relations. In network models, disorders are conceptualized as systems of causally connected symptoms rather than effects of a latent disorder.

Mental disorders cannot be identified independently of their symptoms. In medicine, the medical condition can be separated from the symptoms. In psychology, this is not possible. In order to separate this, it must be possible that a person has symptoms without the disorder (e.g. depression without feeling down is not possible). In mental disorders, it is likely that there is symptom-symptom causation. One symptom causes another symptom and this leads to a mental disorder.

With network systems, it might be unclear where one disorder starts and another stops. The boundaries between disorders become unclear. Network models might change treatment, as the treatment is then no longer aimed at the disorder but rather at the symptoms and the causal relationship between the symptoms.

Networks in psychopathology can be created by using data on symptom endorsement frequencies (e.g. looking at correlations between symptoms) (1), assess the relationship between symptoms rated by clinicians and patients (2) and use the information in the diagnostic systems (3).

In networks, any node can reach another node in only a few steps. This is called the small world property. The DSM attempts to be neutral, theoretically, but makes claims about causal relationships between the disorders.

Asking experts on how nodes are related (e.g. clinicians and symptoms of a disorder) is called perceived causal relations scaling

Extended psychopathology systems refers to network systems in which the network is not isolated in a single individual but spans across multiple individuals. This would mean that one symptom in one person could cause a symptom in another person. These networks can be used to review what the interaction is between symptoms of different people in different social situations.

Association networks show what the strength of the correlations between symptoms is. This gives an indication of different disorders, as the symptoms in disorder A are more correlated with the other symptoms in disorder A than with the symptoms of disorder B.

A partial correlation network, also called a concentration network, shows the partial correlations between symptoms. This can be used to be a bit more certain about the causal relationship of two nodes as it rules out some third-variable explanations. Concentration graphs can be used to assess which pathways between symptoms appear common in a disorder.

Association and concentration graphs provide information about the causal relationship between nodes but it does not provide information about the causal direction of the network. Directed networks give information about the causal relationship between nodes. This is usually represented in a DAG. In order to generate statements about the initiation, maintenance and treatment of disorders of individuals, the network of individuals needs to be researched.

There are three possible causal pathways:

1. Y mediates the causal relationship between X and Z.
2. X is a common cause of Y and Z.
3. Z is a common effect of X and Y.

If network analysis tries to construct disorder networks for individuals, cross-sectional data is not useful. In order to construct individual disorder networks, time-series data may be obtained. This can be done by asking people to report on their symptoms every day for a number of consecutive days.

There are two ways in which a network can harbour risk of developing a certain mental disorder:

1. The network itself might be risky. The strengths of associations can then be analogous to domino tiles.
2. Some symptoms might have a stronger causal influence on the rest of the network for an individual.