The Catalog vs CogPO¶
CogPO, the Cognitive Paradigm Ontology, and the Catalog both name experimental paradigms and both decouple paradigms from the mental processes they are supposed to measure. From there they diverge. The Catalog names standard tasks, tests membership by procedure, and links each task to the processes it engages. CogPO describes a paradigm by its observable components, the stimuli, responses and instructions. The CogPO paradigm is defined with a granularity finer than that defined by the Catalog. Each has what the other lacks.
What is CogPO? provides an overview of CogPO. The task mapping page gives the correspondence between the two paradigm lists row by row, the dimension mapping page sets CogPO’s stimulus, response and instruction vocabularies beside the Catalog’s facets and the HED schema, and the CogPO mapping method page explains how each row was decided. This page says what it adds up to.
Two resources doing different jobs¶
CogPO was built for BrainMap, a database of coordinate-based neuroimaging results. Its job is to let a curator code a published experiment so that experiments can be compared and pooled: which paradigm classes apply, what the participant saw, did and was told in each condition. Hence the “in at least one of the conditions” clause in every definition, the instruction to co-code overlapping classes, and a paradigm list that includes acupuncture, breath holding and chewing beside Stroop and n-back. Anything an imaging study might do had to be codable.
The Catalog was built for HED annotation of event-structured data. Its job is to give a dataset a task name that carries a stable meaning: a procedure that can be recognized, a set of processes that a reader can expect the events to bear on, and a list of the variations that keep the name. Hence the inclusion tests, the variation rules, the references and the process axis, and the exclusion of procedures with no trial structure.
The difference in scale is smaller than with the Cognitive Atlas, because both resources are curated lists rather than open contributions.
The Catalog |
CogPO |
|
|---|---|---|
Experimental paradigms |
103 tasks and 4 pseudo tasks |
83 classes (84 on the wiki) |
Cognitive processes |
172 |
not modelled |
Named variations |
776 |
not modelled |
Stimulus, response and instruction vocabularies |
facets, seeded from CogPO |
6 vocabularies, about 100 values |
References per paradigm |
curated, with roles |
occasional BrainMap citation on a definition |
Last curated |
ongoing |
2011 |
How much overlaps¶
Of the 86 CogPO paradigm classes, 50 have a counterpart in the Catalog (27 exact, 15 close, 8 related) and 36 have none; 5 of the matched classes resolve to a named variation rather than a task.
Tasks in the Catalog, by the match type of their primary CogPO class:
Match type |
Tasks |
Share |
|---|---|---|
exact |
24 |
22% |
close |
22 |
21% |
related |
26 |
24% |
none |
35 |
33% |
CogPO paradigm classes, by the match type of their counterpart in the Catalog:
Match type |
CogPO classes |
Share |
|---|---|---|
exact |
27 |
31% |
close |
15 |
17% |
related |
8 |
9% |
none |
36 |
42% |
The none rows on the CogPO side divide into kinds, and the kind is a hand-checked
judgement here (unlike the rule-derived grouping on the Atlas pages, since 86 classes
can be read one by one):
What the class is |
CogPO classes |
|---|---|
Paradigms the Catalog does not list |
15 |
Broad classes with no single counterpart |
3 |
Physiological procedures and interventions |
11 |
Motor acts |
6 |
Imaging protocol |
1 |
Where CogPO registers a specific case as a class of its own, the Catalog often records it as a named variation: Auditory Oddball and Visual Oddball, Delayed Non Match To Sample, the prosaccade, covert verb generation.
Mapped to |
CogPO classes |
|---|---|
A Catalog task |
45 |
A named variation of a Catalog task |
5 |
Where they agree¶
Both put procedure first. A CogPO paradigm class is defined by what the participant sees and does, never by the construct it is claimed to tap, and the Catalog’s inclusion tests do the same. The 27 exact matches are the paradigms that any list of standard cognitive tasks converges on, and their definitions in the two resources describe the same procedure in different words.
Both split the paradigm from its response modality only when the field does. CogPO has overt and covert classes for naming, reading, repetition, word generation and word stem completion, because in an fMRI scanner the difference between saying a word and thinking it is the difference between two experiments; the Catalog treats the response as a variation. Neither treats the response modality as making a new task anywhere else.
Where they differ¶
Process axis. CogPO has none. The Catalog’s 172 processes and its task-process links have no counterpart, and nothing in CogPO says that the Stroop task engages selective attention and interference control. This is the largest difference and it is by design: CogPO left “what is measured” to BrainMap’s separate behavioral-domain taxonomy.
Scope. 36 CogPO classes have no counterpart in the Catalog. Fifteen are paradigms the Catalog could list and does not: Braille reading, deception, counting or calculation, divided auditory attention, film viewing, motor imagery, visual imagery, music comprehension or production, olfactory discrimination, overt reading, covert and overt recitation, video games, and covert and overt word stem completion. Eleven are physiological procedures or interventions (acupuncture, breath holding, chewing, eating, micturition, sleep, TMS, electrical, thermal, painful and vibrotactile stimulation), six are motor acts (grasping, pointing, drawing, writing, whistling, flexion or extension), three are broad perceptual classes and one is the flashing checkerboard. The Catalog’s task criteria exclude the last four groups deliberately.
In the other direction, 35 of the Catalog’s 107 tasks have no CogPO class at all and 26 more fall only under a broad class such as Reward Task or Visual Attention. CogPO has no economic games, no metacognition, no priming, no spatial navigation, no implicit learning, no complex span, no stop-signal and no questionnaire.
Broad classes. Ten CogPO classes are families rather than paradigms: Encoding,
Episodic Recall, Reward Task, Semantic Discrimination, Visual Distractor/Visual
Attention (CogPO’s own word for it is “catch-all”), Face Monitor or Discrimination and
their kin. Nine Catalog reward tasks fall under Reward Task Paradigm alone. Where the
Catalog has a specific task inside such a class, the row is close; where it has only
neighbours, related.
Variations, inclusion tests, references. CogPO has none of the three.
What CogPO adds¶
The dimension vocabularies. CogPO is the only one of the resources compared on this site
that says, in a closed vocabulary, what a participant is shown, what they do with which
body part, and what they are told to do. The Catalog has needed exactly this since its
facets were first planned, and it adopts CogPO’s six vocabularies as the starting point
for its stimulus, response and instruction facets, crediting every value to CogPO in
data/facet_defs.tsv and marking every addition. The
dimension mapping page lists the values, the additions and the
few values dropped as out of scope.
CogPO and the HED schema¶
CogPO’s dimensions are close kin to the HED schema: each of the six lands in one HED subtree, and 88 of the 102 CogPO values have a HED 8.4.0 tag. That kinship is why the Catalog can adopt CogPO’s vocabularies as facets without inventing a parallel one, and why a facet value on a task can become a tag on the dataset’s events. HED vs CogPO sets the two out in full: what each describes, where the vocabularies meet value by value, and what each has that the other lacks.
What the comparison changes¶
For the Catalog, CogPO is a source of vocabulary more than of paradigms. The fifteen task-shaped classes the Catalog lacks are candidates for the issue tracker, not omissions: most describe imaging-era block designs (film viewing, overt reading, recitation) whose event structure is thin. The dimension vocabularies, by contrast, go straight into the facets.
For a reader of CogPO, the Catalog supplies what CogPO left out: a process layer, variations, inclusion tests and references for the 50 classes the two share, and, for the classes CogPO never populated, the paradigm-to-dimension assignments that its design called for and its curation never reached.