STUDY DESIGN.: Cross-sectional design. OBJECTIVE.: To study the correlation between intramuscular EMG (iEMG) and surface EMG (sEMG) from lumbar multifidus and erector spinae muscles during (submaximal) voluntary contraction tests in healthy participants. SUMMARY OF BACKGROUND DATA.: Low back muscle function is a key component in the stability of the lumbar spine in which an important role is attributed to the lumbar multifidus (LM). Impairments in this stabilization system are held responsible for (chronic) low back pain. LM function can be measured by iEMG and sEMG, however, in earlier studies, results from iEMG and sEMG were inconsistent. METHODS.: Fifteen healthy adults were included. The intervention consisted of five clinical tests: resting, submaximal contraction tests of the lower back, abdominal contraction and a biofeedback test in which LM and erector spinae (ES) activities were compared by iEMG and sEMG. Correlations were calculated with regard to original signal, co-contraction ratio and cross-talk ratio. Correlation coefficients for each combination of iEMG and sEMG signals were calculated, to identify original signal (i.e. activity of only the targeted muscle) and possible cross-talk. Correlations > 0.75 were considered as good concurrent validity. RESULTS.: The original signals of LM showed fair to high correlation coefficients (r: 0.3 – 0.8). Co-contraction of LM and ES was observed during all tests, but iEMG show more variation in the correlations (r: 0.1 – 0.8) compared to sEMG (r: 0.3 – 0.8). Significant cross-talk was observed in all tests, particularly during the biofeedback test of iEMGES versus sEMGLM and iEMGLM versus sEMGES (r = 0.8). CONCLUSION.: Surface EMG of ES and LM are no adequate representation of LM and ES activity measured by iEMG because of moderate/high cross-talk and co-contractions. Clinical tests that aim to assess LM activity do not represent isolated LM activity. This should be taken into account in future clinical studies.Level of Evidence: 3