In-the-ear calibration of sound pressure level may be problematic at frequencies above 2 kHz, because the pressure can vary significantly along the length of the ear canal, due to reflection of sound waves at the eardrum. This issue has been investigated by measuring behavioral thresholds to tones in a group of human subjects (N = 61) for two different insertion depths of an insert earphone. The change in insertion depth was intended to alter the distribution of pressure in the ear canal, shifting the frequency at which spectral notches occur. The inset earphone or "probe" (Etymotic ER-10C) also contained a calibrated microphone, allowing the recording of sound pressure levels in the ear canal. Prior to the threshold measurements in each subject, the Thevenin acoustic source characteristics of the probe were determined by a special calibration procedure. This calibration allowed the expression of the sound level at threshold in terms of acoustic intensity (W/m2). The impact of changes in insertion depth was determined by measuring behavioral threshold at each depth. Because cochlear sensitivity remained constant, the level of sound entering the ear at threshold should have been the same (within measurement error) for both insertions. The difference in sound pressure level (SPL) at threshold between the two probe insertions was greatest at the notch frequency of the first insertion. At this notch frequency, the SPL at threshold increased by an average of 11.4 dB. The change in sound intensity level (SIL) at threshold was almost always less than the change in SPL. At the notch frequency, the SIL decreased, on average, by only 0.5 dB. These results suggest that SIL may be a better indicator than SPL of the sound level entering the ear, especially for frequencies in the 4-8 kHz range.