By analyzing the long-term (1950-2007) variability of diurnal temperature range [DTR; daytime maximum (T_max) minus nighttime minimum (T_min)] at 21 stations in Taiwan, this study applies univariate analysis, a trend-free prewhitening procedure combined with a modified Mann-Kendall test, and EOF-based multivariate trend analysis (TEOFA) to the ranked station DTR, T_max, and T_min. To reveal the large-scale associations with the local TEOFA results, this study also uses global gridded T_max,T_min, nighttime marine air temperature, and sea level pressure (SLP) datasets archived at the U.K. Meteorological Office Hadley Centre. On the basis of the signs and relative magnitudes between the annually-mean T_max and T_min trends, the stations can be classified into three types through univariate analysis. With a common increasing T_min, type-A (type-B) stations have an increasing T_min (Tmax) faster than T_max (T_min) whereas type-C stations have a decreasing T_max. Both type-A and type-C (type-B) thus show(s) a decreasing (an increasing) DTR. For most stations, the increasing T_min is the largest in December-February. In contrast, all type-B stations have an increasing T_max peaking in June-August. Noticeably, six of nine type-B stations, either in the remote islands or in the seaports, are particularly influenced by the ocean. Three DTR trend modes stand out of TEOFA. The first mode, TEOF₁, captures increasing DTR trends at four type-B stations and decreasing DTR trends elsewhere. The associated increasing T_min (T_max) trend is consistent with the increasing (decreasing) clouds during nighttime (daytime) and is well-correlated with the large-scale patterns, suggesting that it is part of the global warming scenario. Nevertheless, TEOF₁ also captures a decreasing T_max trend in Taiwan’s highly developed western plains (i.e., type C) and eastern China where an increasing SLP pattern is observed, implying the anthropogenic forcings on DTR. TEOF₂ (TEOF₃) depicts the decadal-to-interdecadal DTR variability in central (northern) Taiwan. Evolution of TEOF₂ shows smaller (larger) amplitude before 1970s (after mid-1980s). The associated large-scale patterns suggest that TEOF₂ (TEOF₃) captures the relationship between an intensified (weakened) East Asian winter monsoon and La Niña (El Niño)-like condition in eastern (central) equatorial Pacific. Embedded in a Pacific-Japan-like teleconnection pattern, the southwestward intrusion of Pacific subtropical anticyclone in June-August that signifies the weakened southwesterly monsoon is also depicted by TEOF₂.