"We used a 1 kHz calibration tone to set the VU meter on the audiometer and then to check the output from the audiometer we have measured the levels using the a standard coupler-SLM arrangement and TDH 39 headphones connected to the audiometer."
1 kHz tone 60dB (dial) audiometer = 76dBSPL on SLM
speech 60dB (dial) audiometer = 60dBSPL on SLM
Answer: I interpret this to mean you are presenting the 1 kHz reference signal through the speech circuit of the audiometer? In that case then the RETSPL for TDH-39 is 19.5 dB - meaning your measured SPL on the SLM for the 60 dB dial should be 79.5 dB SPL for the 1 kHz pure tone, right? i.e. it seems you’re just a fraction outside the ± 3 dB tolerance with your reference tone.
Aside from that, it seems that maybe your speech level for 60dB HL (dial) is indeed a little low based on the info provided.
IEC 60645-2:1993 Audiometers - Part 2: Equipment for speech audiometry
"We did run the calibration tone through the audiometer’s speech circuit (via Ext A) yes - that is a good point. So does this mean we could add a 3dB correction factor when we use TDH headphones? Or, if we calibrated the VU meter to -3dB would that be similar? How does this work with the speech output which measured at 60 dB SPL with 60dB HL dial on the audiometer? Should the measured output be different for the 1 kHz tone and the speech signal?"
Answer: Well I would say that the most sensible approach would be to go into calibration mode and adjust the audiometer’s output by 3 dB to get the reference tone to within tolerance, rather than use the VU meter.
According to the standard (IEC 60645-2), the specifications and test methods for speech audiometers are “based on the assumption that the calibration signal level of the recorded speech material is the same as the average level of the speech material when measured in a specified manner.”
So I would interpret this to say that in addition to the 3 dB adjustment, the measured output of your speech signal should match the 1 kHz calibration tone.
What that would then mean is that if you sent someone your speech stimuli (e.g. on a CD) and then they put it through their own audiometer they should end up playing the speech stimulus at the same average level as you did, with positive and negative peaks in the speech occurring at the same higher and lower levels accordingly (provided they calibrate the 1 kHz reference signal to the RETSPL level like you (60 dB dial = 79.5 dB SPL).
However, if reference tone and speech level do not match then you just need to state the relationship of the speech level to the reference so that someone else with their audiometer could still play the same speech at the same level as you did.
On this topic, 60645-2 says the following “If the speech and calibration signal are not at the same level the method of calibration shall be described. If the level of the calibration signal and the average level of the speech material are different, calibration and test methods should be modified as recommended by the producer of the speech test material.”
So, there’s really no “should” or “shouldn’t” as to whether the speech and 1 kHz reference tone output should be different; you should either match them or note the difference in order to maintain standardisation for any future studies.
By the way - the use of the 1kHz tone as a reference signal deviates slightly from the standard. I know the technical reasons often is that it gives a stable signal for convenience of measurement in a coupler but it is a discrete frequency and steady state, whereas the speech signal is broadband and fluctuating - 60645-2 says the calibration signal should be a weighted noise, a band of noise or a warble tone centred on 1 kHz with a 1/3 octave bandwidth.