Introduction In practice, inferences on a quantile or several quantiles of a normal population may be more informative than inferences on the mean or variance of the population. Some researchers discussed the problem of making inferences on several quantiles when the underlying population has a normal distribution. They described that constructing a set of simultaneous confidence intervals (SCI) for several normal quantiles may be of interest in medicine or engineering studies (e.g. in the body weight or blood pressure or drug-drug interaction studies or in measuring the reliability of systems). We consider the problem of constructing SCI for all quantiles of a normal population. We propose an exact set of SCI for all normal quantiles. We also propose two algorithms for calculating the required critical values. Then, we illustrate and discuss the performance of the methods with two real examples and a simulation study. Material and Methods Suppose that X1, ... ,Xn1is a random sample of size ݊n ≥ 2 from a normal population with mean μ and variance σ2. Let ܼZp denote the p-th quantile of the standard normal distribution and let qp = μ+Zpσ denote the p-th quantile of a normal population. If ܺx̄ and ܵS2 are the sample mean and variance, q̂p = x̄+aSzp is the uniformly minimum variance unbiased estimator (UMVUE) of qpwhere ܽa = √(n-1)/2Γ((n-1)/2)/Γ(n/2). The UMVUE of vp = var(q̂p) = bpσ2 is ...p = bpS2where bp1/݊ n+Zp2(a2-1). We propose T = maxp∈(0,1) |Tp| where Tp = (q̂p-qp)/...p1/2 as a pivotal quantity for constructing SCI for qp for all p ∈ (0,1). We derived the maximum value of ܶ in a closed form and consequently, we could provide two numerical algorithms to calculate the required critical values of T. Results and discussion Rosenkrantz [3] proposed a conservative set of SCI for qp for all p ∈ (0,1). Rosenkrantz proposed to draw the graph of the empirical quantile along with the corresponding simultaneous upper and lower bounds for these quantiles. He mentioned that “the more the parent distribution of the data departs from the normal, the more likely the empirical quantile plot will have points that lie outside these bounds.” Although the proposed bands by Rosenkrantz are easily computable, they are not optimal (they are conservative). Our proposed SCI are exact and thus these bands are narrower than those of Rosenkrantz. Therefore, our proposed set of SCI performs better than the one proposed by Rosenkrantz in terms of the volume of the SCI. Liu et al. [4] also proposed SCI for several normal quantiles. Although the proposed SCI by Liu et al. is exact, they are not for all normal quantiles. Our proposed SCI is also applicable for several normal quantiles and in this case, our SCI will be conservative. However, the performance of our SCI is comparable with that of Liu et al. when the number of considered quantiles and sample sizes are increased. Our findings are valid when the underlying population is normally distributed; however, these results can be extended to the log-normal populations. When the population distribution is not known, constructing a set of non-parametric SCI for all quantiles is a complicated task. One can refer to [4] for more discussion. Conclusion We proposed an exact set of SCI for all normal quantiles. We could derive a closed form of the pivotal quantity and its distribution function. Therefore, we could calculate the required critical values. Our proposed SCI performs better than or comparable with the existing methods in terms of the volume of SCI. Therefore, our proposed SCI can be recommended in practice for making simultaneous inferences on several normal quantiles. [ABSTRACT FROM AUTHOR]