Abstract <p>A brief analysis of publications addressing the influence of acoustic coupling on signals recorded during ultrasonic testing of components and welded joints is provided. It is noted that most often the authors pay attention to the amplitudes of pulses, but do not analyze their spectrum and its alteration in relation to the quality of preparation of contact surfaces. Experimental studies of backwall echoes received from carbon and austenitic steel specimens have demonstrated that variations in surface roughness (within the range of Rz 10 to Rz 80 μm) not only alter the amplitude of the detected signals but also cause significant distortions in their waveform and spectral composition. In some cases, the operational frequency of the backwall signal may decrease by 30% or more relative to the nominal frequency of the piezoelectric transducer. Measurements performed on specimens with both deterministic and stochastic surface roughness profiles have shown that the quality of the coupling surface preparation has a substantially greater impact on the spectral characteristics of the received signals compared to the backwall surface roughness. The findings highlight the necessity of accounting for the influence of coupling surface roughness in test objects and welded joints on the spectral content of ultrasonic signals when developing procedural guidelines for pulse–echo and diffraction-based ultrasonic testing methods.</p>