# Help needed with analyzing sensitivity to changes in Musical sounds

#### icsgreen

##### New Member
Thank you so much for your advice in advance, I owe you a pint if you are in London!

I am trying to analyze whether kids with autism are more sensitive to changes in timbre (tone color) than typical kids. Timbre research shows that we hear timbre in 3 dimensions: attack time (ATT), overall balance of high and low harmonics (SCG), and ratio of even to odd harmonics (EHA). Here's a helpful image:

For this experiment there were 3 tones along each of the seven dimensions dimensions shown (1D, 2D, 3D), plus a standard tone in the corner. The standard tone had a value of zero for all three dimensions (0_0_0) and, for example, the third tone along the 3D axis had the value 3_3_3. All changes were perceptually equal based on previous research.

Participants heard 2 tones and had to say whether they were the same or different. I want to analyze how sensitive each participant was to changes along a given dimension (1D, 2D, and 3D). Here is the problem that I am running into:

The pairs of stimuli that each participant heard did not all include the standard nor did they all fall along the same axis. For example one of the stimulus pairs might have been 1_0_0 and the next 2_2_0 (ATT_SCG_EHA). Therefore if a child detected a change here, it seems impossible to say whether they detected the change is attack of 1 degree, the change in Spectral Center of Gravity of 2 degrees, or the interaction of the two.

If I had included the standard in each pair or made each pair fall along an axis (1D, 2D, or 3D) then I would have a lot more to work with. What I have done so far is to calculate an overall d' (signal detection) for all pairs containing changes along ATT, SCG, and EHA separately. I have also calculated percent correct (weighted for degree of change) for all pairs that include the standard or fall along an axis, but there weren't very many of these.

Thank you sooooo much, and do let me know if you have further questions!

Cheers!

#### victorxstc

##### Pirate
Hi :welcome: and thank you for this interesting question.

The experts would come sooon to help you! However, before that, it would be very good if you could organize your dependent variable(s): hearing sensitivity; what else? do you consider ATT, SCG, and EHA "separate" dependent variables or combine them into a composite scale? what else?, the independent variables (the tones in different dimensions and their compositions; what else?), your sample sizes in each group, the number of groups, are your experiments repeated for each single subject? what was your control group? normal children? Or the extent of the change in musical tones? Did the sequences of tones matter and standardized? were the ups and downs random? Did you include any placebo tones between the real tones to assess the attention of children and rule out random responses? Also tell us more about the nature of your variables. For example is EHA continuous or ordinal?

besides these are too technical at least for me, and need more comprehensive explanations:

The pairs of stimuli that each participant heard did not all include the standard nor did they all fall along the same axis. For example one of the stimulus pairs might have been 1_0_0 and the next 2_2_0 (ATT_SCG_EHA). Therefore if a child detected a change here, it seems impossible to say whether they detected the change is attack of 1 degree, the change in Spectral Center of Gravity of 2 degrees, or the interaction of the two.

If I had included the standard in each pair or made each pair fall along an axis (1D, 2D, or 3D) then I would have a lot more to work with. What I have done so far is to calculate an overall d' (signal detection) for all pairs containing changes along ATT, SCG, and EHA separately. I have also calculated percent correct (weighted for degree of change) for all pairs that include the standard or fall along an axis, but there weren't very many of these.
Moreover, I couldn't open the image and see the dimensions. Maybe their explanation might help too.