TLE3, a nuclear transcriptional repressor downstream of WNT signalling pathways, has been hypothesised as predictive of benefit from adjuvant taxane. MA.21 tissue microarrays were constructed from 1097 out of 2104 (52%) patients. TLE3 staining by immunohistochemistry used validated methodology. Continuous TLE3+ (percentage of cells staining positive) was assessed with both visual and automated scoring. The primary objective was to test the predictive effect of TLE3 on relapse-free survival using the MA.21 EC/T and CEF arms and the previously defined cut-point of 30% of cells staining positive in ⩾1 core/tumour. MA.21 patients had 83.2% TLE3 positive (TLE3+) tumours by visual score and 80.6% TLE3+ by automated image analysis while the previously observed rate of TLE3+ cases was 58.6%. TLE3 expression was significantly associated with ER expression (91.2% of ER-positive tumours were TLE3+; P<0.0001). At median 8-year follow-up, there was no evidence of a predictive effect of TLE3 expression with respect to taxane benefit using the established 30% or exploratory quartile cut-points. Proportionately more MA.21 patient tumours than expected were TLE3+. The pre-specified TLE3+ cut-point of 30% was not predictive of taxane benefit. TLE3 expression does not represent a viable biomarker for taxane benefit in breast cancer.