Laboratory assessment of thyroid function is now often initiated with a low pre-test probability, by clinicians who may not have a detailed knowledge of current methodology or testing strategies. Skilled laboratory staff can significantly enhance the choice of appropriate tests and the accuracy of clinical response; such involvement requires both appropriate training and relevant information from the clinician. Measurement of the serum thyroid stimulating hormone (TSH) concentration with an assay of adequate sensitivity is now the cornerstone of thyroid function testing; for untreated populations at risk of primary thyroid dysfunction, a normal TSH concentration rules out an abnormality with a high degree of certainty. However, in several important situations, most notably pituitary abnormalities and early treatment of thyroid dysfunction, serum TSH can give a misleading indication of thyroid status. An abnormal TSH concentration alone is never an adequate basis for initiation of treatment, which should be based on the typical relationship between trophic and target gland hormones, based on serum TSH and an estimate of serum free thyroxine (T4). Six basic assumptions, some clinical, some laboratory-based, need to be considered, together with the relevant limiting conditions, for reliable use of this relationship. Current methods of free T4 estimation remain imperfect, especially during critical illness. Diagnostic approach differs significantly between initial diagnosis and follow-up of treated thyroid dysfunction. In some situations, serum triiodothyronine (T3) is also required, but serum T3 lacks sensitivity for diagnosis of hypothyroidism, and has poor specificity during non-thyroidal illness. Where assay results are anomalous, most atypical findings can be resolved by attention to the clinical context, without further investigation.