Instrumentation
Trial-level events captured choice, condition, response time, and participant context.
DIEGOCORDOVA / CASE STUDY 01MEMORY CAPACITY RESEARCH · 2026
Cognitive science × product analytics
How precise is visual memory?
A browser experiment reveals a clear descriptive gradient: we remember objects well, but their exact state is easier to confuse.
RESEARCHERJavaScript + jsPsychR + mixed effects
Results overview
41
Participants
3,116
Test trials
86.8%
Weighted accuracy
3
Conditions
01 / Behavioral results
Similarity changed the difficulty of remembering.
Compare conditions
Performance profile
Text alternative: Accuracy by condition. Novel 95.5 percent, Exemplar 88.8 percent, State 80.8 percent. Chance is 50 percent. Results are descriptive summaries from the processed dataset.
02 / Experiment design
A controlled path from exposure to choice.
Stage 01
Encode the object
Participants studied object images in a randomized sequence, with pair membership counterbalanced across the experiment.
03 / Research copilot
Interrogate the evidence—not the imagination.
Memory Analyst
Demo analysis · evidence locked
Ask about the study design, metrics, findings, or next steps. Answers stay within the verified case-study evidence.
Evidence-grounded response
The clearest descriptive pattern is a precision gradient: Novel foils were recognized most accurately and quickly (95.5%, 1,252 ms), Exemplar was intermediate (88.8%, 1,428 ms), and State was hardest (80.8%, 1,530 ms). No p-values are available here, so this is an observation—not a significance claim.
04 / Analysis pipeline
From interaction events to interpretable evidence.
- 01
Capture
jsPsych trial data
- 02
Store
CSV / OSF storage
- 03
Clean
R joins + JSON extraction
- 04
Summarize
Participant × condition metrics
- 05
Model
Mixed effects + ggplot
Methods note
Trial-level outputs and participant metadata were cleaned and joined with tidyverse, readr, and jsonlite. Accuracy and reaction time were summarized by participant and condition; lme4 and lmerTest supported mixed-effects modeling. The numbers presented here are descriptive summaries from the processed dataset.
05 / Product translation
The research craft maps directly to product decisions.
Metric definition
Accuracy and median response time separated outcome quality from interaction effort.
Experimental controls
Randomization and counterbalancing reduced avoidable order, position, and assignment bias.
Segmentation
Condition-level cuts revealed where visually similar alternatives changed behavior.
Behavioral analysis
Within-participant structure connected individual behavior to aggregate patterns.
Decision communication
A precise narrative distinguishes observed evidence from proposed explanations.
06 / Reflection
What I would do next.
- 01Preregister directional hypotheses before collection.
- 02Define exclusion rules for attention checks in advance.
- 03Report confidence intervals and mixed-model coefficients.
- 04Test the pattern with a larger, more diverse sample.
07 / Try it yourself