Fig 1.
Research map of the results in a published article [8].
Each node in a research map has three properties: What (top), Where (middle), and When (bottom). Nodes are connected by edges that represent relations: Excitatory (sharp arrowhead), Inhibitory (blunt arrowhead), and No-connection (dotted line, circular arrowhead). Each empirical edge also has a score that reflects the amount of evidence represented, as well as symbols that reflect the experiment classes recorded for that edge. Scores and experiment symbols are not assigned to hypothetical edges. Users can highlight edges that reflect the main idea(s) discussed in the article, so that they are more apparent. In cases where no one relation has received dominant evidence, the corresponding edge is represented by a diamond arrowhead and is not assigned a score.
Fig 2.
A shorthand method for calculating the score of an edge in a research map.
A table representing the model space of experiments is instantiated with a pseudocount of one (a form of Laplace smoothing). The symbols along the left indicate the classes of experiments involving an Agent, A: Positive Intervention (A ↑), Positive Non-intervention (A⌀↑), Negative Non-intervention (A⌀↓), and Negative Intervention (A ↓). The symbols along the top indicate the results recorded in a Target, B: increase (B+), no change (B0), and decrease (B−). This particular instantiation of the scoring table encodes four (5 − 1) Positive Interventions that caused the Target to increase, one (2 − 1) Positive Non-intervention that caused the Target to decrease, and one (2 − 1) Negative Non-intervention that caused the Target to decrease. There are thus five experiments suggesting an Excitatory relation (green regions), and one experiment suggesting an Inhibitory relation (red region).
Fig 3.
The growth of an edge’s score due only to consistency (left) and due to convergence (right).
These plots show how the score of a research-map edge increases with each subsequent experiment (all with agreeing results), due to the principle of consistency (left) and due to the principle of convergence (right). The plot on the left represents repeated iterations of the same class of experiment (e.g., Positive Intervention) with consistent results. The plot on the right represents multiple iterations of experiments in which, at each iteration, one of the least-represented classes of experiments was performed, leading to consistent results. These two plots express an axiom of research maps: the principle of convergence carries greater epistemological weight than the principle of consistency.
Fig 4.
An example of an edge in a research map.
This research map encodes three experiments—two Positive Interventions (↑) and one Negative Intervention (↓)—involving CREB and the number of Arc neurons. This map is part of a larger one that is discussed below.
Fig 5.
Using hypothetical edges to organize research maps.
The example above shows how hypothetical edges (in gray) help to organize empirical edges in a research map, thus framing the empirical results in light of a specific hypothesis.
Fig 6.
Entering information in ResearchMaps.
The left panel shows the interface used to input information. The citation on the top refers to the article whose research map is displayed. Highlighted in yellow are the edges that reflect the main findings in that article. Users can double-click on any edge in the research map to retrieve PubMed citations that are potentially relevant to the edge’s Agent–Target relation.
Fig 7.
Interacting with the information in ResearchMaps.
This screenshot shows the interface used to interact with the information in the app. The panel on the left is used for entering the details for a particular query (e.g., CREB). The map shown on the right includes only a fraction of the edges that this query returned. This map represents integrated data from many different research articles.
Fig 8.
Establishing the provenance of edges in the Global Map page.
The table in the bottom right of the screenshot appeared in response to clicking on the edge above the table. Hyperlinks in the left column of the table direct users to the individual research maps for each empirical and hypothetical assertion represented in that edge.
Fig 9.
Initial map of experiments exploring the role of CREB in amygdala memory enhancements.
This research map represents a series of experiments designed to explore the role of CREB expressed in a subset of lateral amygdala neurons in an enhancement of auditory and contextual conditioning.
Fig 10.
Map of experiments exploring the role of CREB in memory allocation.
Inventing the concept of memory allocation with the help of research maps not only helped to structure our experiments to test the role of CREB in the amygdala during memory formation but also helped us to plan future experiments. The edge highlighted in orange points to the key experiments in the map representing early experiments on memory allocation. All the other edges represent control experiments that helped us to interpret the memory allocation experiments.
Fig 11.
Global research map of experiments in memory allocation and other related work.
This is a personally curated research map of work in the field of memory allocation and other related work that either overlaps or connects to the work in memory allocation. To minimize the number of nodes, only the What property of each node is shown, so that nodes with different Where and When properties (but identical What properties) are collapsed into one. Nodes in orange appear only in research maps for articles on memory allocation. Nodes in red appear not only in research maps for articles on memory allocation but also in research maps for related work. This research map has helped us to contextualize work in memory allocation and propose hypotheses concerning the mechanistic basis of this process.
Fig 12.
Connectivity characteristics in the Global Map for memory allocation.
The plot shows the surprisingly extensive connectivity between papers in memory allocation and other related work, which can be visualized in the Global Map of ResearchMaps. For example, the graph shows that within three edge traversals, there are over 500 nodes that connect with nodes in individual memory allocation research maps. This extensive connectivity provides abundant opportunities for hypothesis building, since any one of the connected nodes could modulate unknown features of memory allocation (and vice versa).
Fig 13.
An example of how the structural information in a research map can help to identify conflicting results.
The edges in red are in conflict with the edges in black—all the edges cannot be true simultaneously. (For simplicity, the scores and experiment symbols in this research map have been omitted).