A 2026 study introduced CANID, a three-task battery designed to measure parts of canine memory and executive function in one visit and across repeat visits. The useful finding is methodological: a research measure can show consistent group-level patterns while still requiring careful attention to age, practice and context. It is not a home diagnosis.
Short answer
Yes, the CANID battery was repeatable enough to support longitudinal research in this sample, but the result needs context. Overall scores showed an age-adjusted intraclass correlation coefficient of 0.45, while the baseline-to-six-month and baseline-to-twelve-month comparisons were 0.63 and 0.59. Those are measures of consistency under a particular protocol. They are not diagnostic accuracy, a ranking of dogs, or a stable label for an individual pet.
What the study was trying to measure
Hargrave and colleagues published the paper in Animal Cognition as an Article in Press on October 12, 2026. CANID was built for research on cognitive aging in community dogs. The battery combined three tasks in a single visit lasting about 40 to 60 minutes:
- Spatial Reversal: a dog learns which route around a barrier works, then must adjust when the open route changes.
- Delayed Search: a dog remembers where a hidden food reward was placed after delays and distractions.
- Cylinder: a dog must inhibit a direct reach toward food and use the accessible end of a transparent cylinder.
The researchers also used a brief sensory screen, recorded sessions on video and had a second observer code at least 20 percent of trials to assess agreement. That design matters because a score is only useful when the task, timing and scoring rules are visible.
Who took part, and how often?
The baseline sample included 239 community dogs recruited at three sites in Arizona and California. Dogs were approximately 1 to 16 years old at recruitment. A smaller group returned for repeat testing: 138 dogs contributed six-month data and 114 contributed twelve-month data, with 91 dogs represented at all three timepoints.
The follow-up groups should not be read as if they were the same as the baseline sample. Some dogs were unavailable because of mortality, relocation, owner non-response, scheduling or site logistics. The researchers also stopped a task when a dog became fearful or anxious and could not be soothed, and did not collect data when a dog could not acclimate or accept food rewards.
What the reported numbers show
| Measure | Reported result | How to read it |
|---|---|---|
| Age association | Age accounted for 27% of variance in the overall score | The tested sample showed age-related differences in performance. |
| Age-adjusted repeatability | ICC 0.45, 95% CI 0.29 to 0.58 | Consistency remained moderate after accounting for age. |
| Baseline to six months | ICC 0.63, 95% CI 0.51 to 0.72 | The same battery gave substantially consistent scores across this interval. |
| Baseline to twelve months | ICC 0.59, 95% CI 0.47 to 0.70 | Consistency was also present across the longer interval. |
| Overall retest change | +0.23 SD at six months; +0.04 SD at twelve months | Short-term improvement was visible at six months but not at twelve months. |
Evidence-led insights
1. Repeatability is not the same as diagnosis
Reported: the battery produced repeatable scores across the study intervals. Inferred: that makes CANID useful for asking whether groups or populations change over time. It does not establish that a single result can identify cognitive impairment in one dog. The paper says future work is needed to define within-dog change and a minimum detectable change for individual outcomes.
2. A retest can include learning
The overall score improved by 0.23 standard deviations at six months, but that improvement was not present at twelve months. The task-level pattern was mixed: Spatial Reversal improved at six months, Delayed Search was lower at twelve months, and Cylinder performance showed a small increase that persisted. This is why the authors discuss practice effects alongside possible age-related change. A changed score needs its task and retest interval beside it.
3. Context is part of the measurement
Test location was associated with overall performance after the model accounted for age and other covariates. Dogs tested at Canine Companions scored higher than dogs at the Arizona Canine Cognition Center in the reported model. The authors discuss possible explanations including training history, arousal, temperament and group composition. The result does not identify one cause and does not rank facilities or training programs.
What this study does not show
- It does not provide a home test, a clinical threshold or a diagnosis.
- It does not measure intelligence, emotion, quality of life or a fixed personality trait with one score.
- It does not show that age caused every difference, or that a single task change proves decline.
- It does not make the 239-dog sample representative of every pet dog or every breed.
- It does not rank food, toys, training programs or products.
A careful takeaway for pet families
The practical lesson is about interpretation. Researchers made the question stronger by repeating the same battery, tracking the interval between visits, modelling age and reporting where dogs were tested. The same habits help readers resist over-reading a single observation: keep the context, look for patterns across ordinary routines and avoid turning one behavior into a label. If a dog's daily behavior changes, this paper does not supply an action threshold; it only explains why research teams need repeated, carefully described measurements.
Source note and limits
This is a single-study Methods Note, not a systematic review and not a VEP-owned dataset. The paper was an Article in Press when accessed on October 30, 2026, and the authors say the final edited version may replace it. The study used three United States sites, had follow-up attrition and reported a conflict of interest for one author. The authors identify a public methods manual, video library, data tools and R code in the OSF repository. Read the publisher record and DOI record for the full methods, tables and future research boundaries.
