Limits...
Modelling of human low frequency sound localization acuity demonstrates dominance of spatial variation of interaural time difference and suggests uniform just-noticeable differences in interaural time difference.

Smith RC, Price SR - PLoS ONE (2014)

Bottom Line: This allowed us to model ITD variation for previously published experimental acuity data and determine the distribution of just-noticeable differences in ITD.Our results suggest that the best-fit model is one whereby just-noticeable differences in ITDs are identified with uniform or close to uniform sensitivity across the physiological range.We discuss how our results have several implications for neural ITD processing in different species as well as development of the auditory system.

View Article: PubMed Central - PubMed

Affiliation: Department of Cell and Developmental Biology, University College London, London, United Kingdom ; Centre for Mathematics, Physics and Engineering in the Life Sciences and Experimental Biology (CoMPLEX), University College London, London, United Kingdom.

ABSTRACT
Sound source localization is critical to animal survival and for identification of auditory objects. We investigated the acuity with which humans localize low frequency, pure tone sounds using timing differences between the ears. These small differences in time, known as interaural time differences or ITDs, are identified in a manner that allows localization acuity of around 1° at the midline. Acuity, a relative measure of localization ability, displays a non-linear variation as sound sources are positioned more laterally. All species studied localize sounds best at the midline and progressively worse as the sound is located out towards the side. To understand why sound localization displays this variation with azimuthal angle, we took a first-principles, systemic, analytical approach to model localization acuity. We calculated how ITDs vary with sound frequency, head size and sound source location for humans. This allowed us to model ITD variation for previously published experimental acuity data and determine the distribution of just-noticeable differences in ITD. Our results suggest that the best-fit model is one whereby just-noticeable differences in ITDs are identified with uniform or close to uniform sensitivity across the physiological range. We discuss how our results have several implications for neural ITD processing in different species as well as development of the auditory system.

Show MeSH
Best-fit acuity curves for uniform or linear ΔITD distributions.Candidate uniform and linear ΔITD models used to find best-fit acuity (Δθ) distributions for five acuity data sets. Uniform distributions are described by parameter c0 and linear distributions are described by parameters cp and kp. As with the best-fit descriptions of ΔITD distributions, best-fit acuity distributions have low kp values for the linear ΔITD case, close to the uniform case. AICc values indicate that the majority of acuity data sets are more appropriately described by uniform ΔITD distribution as these have lower AICc values than for a linear ΔITD distribution (comparing the same data sets).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3928360&req=5

pone-0089033-g004: Best-fit acuity curves for uniform or linear ΔITD distributions.Candidate uniform and linear ΔITD models used to find best-fit acuity (Δθ) distributions for five acuity data sets. Uniform distributions are described by parameter c0 and linear distributions are described by parameters cp and kp. As with the best-fit descriptions of ΔITD distributions, best-fit acuity distributions have low kp values for the linear ΔITD case, close to the uniform case. AICc values indicate that the majority of acuity data sets are more appropriately described by uniform ΔITD distribution as these have lower AICc values than for a linear ΔITD distribution (comparing the same data sets).

Mentions: Both uniform and linear ΔITD candidate distributions were tested against the original acuity data. Best-fit acuity curves for uniform or linear distributions are shown in Figure 4 along with their parameters and corrected Akaike information criterion values (AICc). Our predicted acuity values rise steeply towards infinity around 90°, where ITD is maximum (our idealised acuity is discontinuous at ITDmax). Again, we found that the best-fit linear jnd ITD distributions have low proportionality constants and are essentially close to the uniform case. AICc values were used to evaluate how well the non-linear acuity models account for the data. The value of AICc is used to compare different models for the same data set, a lower AICc value indicates a better explanation of the data by the model. The AICc calculations take into account the the number of parameters in a model, thus including a measure of model complexity in order to prevent overfitting of data. We used the corrected version of AIC owing to the low number of data points per data set. We found that all of the data sets except one (Mills figure 5) have lower AICc values for the uniform ΔITD best-fit curves than for the linear ΔITD best-fit curves. This indicates that uniform ΔITD is a slightly more appropriate model for ΔITD variation than linear ΔITD as it accounts for the data without introducing unnecessary complexity. Linear ΔITD best-fit curves also have very low proportionality constants, essentially making them close to the uniform model.


Modelling of human low frequency sound localization acuity demonstrates dominance of spatial variation of interaural time difference and suggests uniform just-noticeable differences in interaural time difference.

Smith RC, Price SR - PLoS ONE (2014)

Best-fit acuity curves for uniform or linear ΔITD distributions.Candidate uniform and linear ΔITD models used to find best-fit acuity (Δθ) distributions for five acuity data sets. Uniform distributions are described by parameter c0 and linear distributions are described by parameters cp and kp. As with the best-fit descriptions of ΔITD distributions, best-fit acuity distributions have low kp values for the linear ΔITD case, close to the uniform case. AICc values indicate that the majority of acuity data sets are more appropriately described by uniform ΔITD distribution as these have lower AICc values than for a linear ΔITD distribution (comparing the same data sets).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3928360&req=5

pone-0089033-g004: Best-fit acuity curves for uniform or linear ΔITD distributions.Candidate uniform and linear ΔITD models used to find best-fit acuity (Δθ) distributions for five acuity data sets. Uniform distributions are described by parameter c0 and linear distributions are described by parameters cp and kp. As with the best-fit descriptions of ΔITD distributions, best-fit acuity distributions have low kp values for the linear ΔITD case, close to the uniform case. AICc values indicate that the majority of acuity data sets are more appropriately described by uniform ΔITD distribution as these have lower AICc values than for a linear ΔITD distribution (comparing the same data sets).
Mentions: Both uniform and linear ΔITD candidate distributions were tested against the original acuity data. Best-fit acuity curves for uniform or linear distributions are shown in Figure 4 along with their parameters and corrected Akaike information criterion values (AICc). Our predicted acuity values rise steeply towards infinity around 90°, where ITD is maximum (our idealised acuity is discontinuous at ITDmax). Again, we found that the best-fit linear jnd ITD distributions have low proportionality constants and are essentially close to the uniform case. AICc values were used to evaluate how well the non-linear acuity models account for the data. The value of AICc is used to compare different models for the same data set, a lower AICc value indicates a better explanation of the data by the model. The AICc calculations take into account the the number of parameters in a model, thus including a measure of model complexity in order to prevent overfitting of data. We used the corrected version of AIC owing to the low number of data points per data set. We found that all of the data sets except one (Mills figure 5) have lower AICc values for the uniform ΔITD best-fit curves than for the linear ΔITD best-fit curves. This indicates that uniform ΔITD is a slightly more appropriate model for ΔITD variation than linear ΔITD as it accounts for the data without introducing unnecessary complexity. Linear ΔITD best-fit curves also have very low proportionality constants, essentially making them close to the uniform model.

Bottom Line: This allowed us to model ITD variation for previously published experimental acuity data and determine the distribution of just-noticeable differences in ITD.Our results suggest that the best-fit model is one whereby just-noticeable differences in ITDs are identified with uniform or close to uniform sensitivity across the physiological range.We discuss how our results have several implications for neural ITD processing in different species as well as development of the auditory system.

View Article: PubMed Central - PubMed

Affiliation: Department of Cell and Developmental Biology, University College London, London, United Kingdom ; Centre for Mathematics, Physics and Engineering in the Life Sciences and Experimental Biology (CoMPLEX), University College London, London, United Kingdom.

ABSTRACT
Sound source localization is critical to animal survival and for identification of auditory objects. We investigated the acuity with which humans localize low frequency, pure tone sounds using timing differences between the ears. These small differences in time, known as interaural time differences or ITDs, are identified in a manner that allows localization acuity of around 1° at the midline. Acuity, a relative measure of localization ability, displays a non-linear variation as sound sources are positioned more laterally. All species studied localize sounds best at the midline and progressively worse as the sound is located out towards the side. To understand why sound localization displays this variation with azimuthal angle, we took a first-principles, systemic, analytical approach to model localization acuity. We calculated how ITDs vary with sound frequency, head size and sound source location for humans. This allowed us to model ITD variation for previously published experimental acuity data and determine the distribution of just-noticeable differences in ITD. Our results suggest that the best-fit model is one whereby just-noticeable differences in ITDs are identified with uniform or close to uniform sensitivity across the physiological range. We discuss how our results have several implications for neural ITD processing in different species as well as development of the auditory system.

Show MeSH