Mortality in head and neck squamous cell carcinoma (HNSCC) is high due to emergence of therapy resistance which results in local and regional recurrences that may have their origin in resistant cancer stem cells (CSCs) or cells with an epithelial-mesenchymal transition (EMT) phenotype. In the present study, we investigate the possibility of using the cell surface expression of CD44 and epidermal growth factor receptor (EGFR), both of which have been used as stem cell markers, to identify subpopulations within HNSCC cell lines that differ with respect to phenotype and treatment sensitivity. Three subpopulations, consisting of CD44(high)/EGFR(low), CD44(high)/EGFR(high) and CD44(low) cells, respectively, were collected by fluorescence-activated cell sorting. The CD44(high)/EGFR(low) population showed a spindle-shaped EMT-like morphology, while the CD44(low) population was dominated by cobblestone-shaped cells. The CD44(high)/EGFR(low) population was enriched with cells in G0/G1 and showed a relatively low proliferation rate and a high plating efficiency. Using a real time PCR array, 27 genes, of which 14 were related to an EMT phenotype and two with stemness, were found to be differentially expressed in CD44(high)/EGFR(low) cells in comparison to CD44(low) cells. Moreover, CD44(high)/EGFR(low) cells showed a low sensitivity to radiation, cisplatin, cetuximab and gefitinib, and a high sensitivity to dasatinib relative to its CD44(high)/EGFR(high) and CD44(low) counterparts. In conclusion, our results show that the combination of CD44 (high) and EGFR (low) cell surface expression can be used to identify a treatment resistant subpopulation with an EMT phenotype in HNSCC cell lines.