Numerous epidemiological studies have shown stronger associations between $PM_{2.5}$ and both mortality and morbidity than $PM_{10}$. The association of $PM_{2.5}$ with respiratory mortality was examined in Seoul, during the period of $1996{sim}2002$. Because $PM_{2.5}$ data were available for only 10% of this time period, a prediction regression model was developed to estimate $PM_{2.5}$ concentration. Death count due to respiratory-related diseases(total respiratory mortality; ICD-10, J00-J98) and death counts(cause-specific mortality) due to pneumonia(ICD-10, J12-J18), COPD(ICD-10, J40-J44) and asthma(ICD-10, J45-J46) were considered in this study. Averaged daily mortality was 5.6 for total respiratory mortality and 1.1 to 1.6 for cause-specific mortality. Generalized additive Poisson models controlling for confounders were used to evaluate the acute effects of particle exposures on total respiratory mortality and cause-specific mortality. An IQR increase in 5-day moving average of $PM_{2.5}(22.6{mu}g/m^3)$ was associated with an 8.2%(95% CI: 4.5 to 12.1%) increase in total respiratory mortality The association of $PM_{2.5}$ was stronger for the elderly ($geq$65 years old, 10.1%, 95% CI: 5.8 to 14.5%) and for males(8.9%, 95% CI: 2.1 to 11.3%). A $10{mu}g/m^3$ increase in 5-day moving average of $PM_{2.5}$ was strongly associated with total respiratory mortality in winter(9.5%, 95% CI: 6.6 to 12.4%), followed by spring(3.1%, 95% CI: -1.2 to 7.5%), which was a different pattern with the finding in North American cities. However, our results are generally consistent with those observed in recent epidemiological studies, and suggest that $PM_{2.5}$ has a stronger effect on respiratory mortality in Seoul.