Specificity
Specificity (True Negative Rate) · Also known as: True Negative Rate, TNR
Specificity measures the proportion of actual negative cases that were correctly identified as negative by the classifier. It answers the question: 'Of all the cases that were truly negative, how many did we correctly reject?' Specificity is complementary to recall and is essential when false positives are costly.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use specificity when false positives are costly or lead to unnecessary downstream actions. In medical screening, high specificity ensures healthy people are not misdiagnosed as sick, avoiding unnecessary treatment. In spam detection, high specificity ensures legitimate emails are not filtered. Specificity is less critical when false negatives (missing positive cases) are the primary concern.
Strengths & limitations
- Directly measures the ability to correctly identify negative cases; high specificity means low false positive rate.
- Interpretable: specificity = 1.0 means no false positives.
- Essential for applications where false positives lead to costly downstream decisions.
- Complements recall and precision to give a complete picture of classification performance.
- Ignores false negatives; a model could have high specificity while missing many positive cases.
- Can be artificially high if the model is overly conservative and rarely predicts positive.
- Does not account for class imbalance or the relative costs of different error types.
- Alone, specificity is incomplete without considering recall.
Frequently asked
What is the relationship between specificity and precision?
Specificity and precision both measure aspects of negative prediction quality, but they differ: specificity measures how well we identify actual negatives (true negative rate), while precision measures how often our positive predictions are correct. They answer different questions: 'Can we trust negative cases?' vs. 'Can we trust positive predictions?'
How does specificity relate to recall and balanced accuracy?
Recall and specificity are symmetric metrics: recall measures sensitivity to positive cases (true positive rate), while specificity measures sensitivity to negative cases (true negative rate). In binary classification, balanced accuracy is the average of recall and specificity: (recall + specificity) / 2, which weights both types of correct predictions equally.
Why might a model have high specificity but low recall?
A model with high specificity but low recall correctly identifies most negative cases (few false positives) but misses many positive cases (many false negatives). This occurs when the model is overly conservative, predicting positive only in cases where it is very confident. This is appropriate only when false positives are extremely costly and missing cases is acceptable.
Sources
- Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27(8), 861-874. DOI: 10.1016/j.patrec.2005.10.010 ↗
- Powers, D. M. (2011). Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation. Journal of Machine Learning Technologies, 2(1), 37-63. link ↗
How to cite this page
ScholarGate. (2026, June 3). Specificity (True Negative Rate). ScholarGate. https://scholargate.app/en/model-evaluation/specificity
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Balanced AccuracyModel Evaluation↔ compare
- F1-ScoreModel Evaluation↔ compare
- Matthews Correlation CoefficientModel Evaluation↔ compare
- PrecisionModel Evaluation↔ compare
- Recall (Sensitivity)Model Evaluation↔ compare