Analysis of stable isotopes in tree rings is an important tool for global change research. There are two major advantages of using tree ring stable isotope data; they can be used for paleoclimate reconstructions with perfect annual resolution, and the confidence limits can be statistically defined. In addition, trees grow worldwide, and so it is possible to examine geographical climatic variations that have occurred in the past. Another advantage of analyses based on isotope ratios in tree rings is that the physiological mechanisms controlling their variations are reasonably well understood, and are relatively simple compared with the numerous factors that control the annual growth increment. Most research on stable isotopes in tree rings has focused on the stable isotopes of carbon, because they are the easiest to measure and show the most rapid development. However, to date, there is no consensus on which component of tree rings best reflects the climatic and environmental changes in terms of the carbon isotope ratio (δ¹³C). In this study, we investigated the differences in δ¹³C among different tissue components of tree rings. Two tree disks of Pinus sylvestris var. mongolica (SZX01-08 and BZ4-10-1.2) were sampled from forest on the north slope of Yilehuli Mountain, Greater Khingan (approx. 51°57'-52°00'N, 124°13'-124°36'E). This site is located in the exclusive cool temperature zone of China that is dominated by coniferous forest vegetation. Tree-ring samples of earlywood (EW) and latewood (LW) were obtained with different sculpturing and pooling programs to avoid interference from non-uniform aspect distributions, i.e. both samples were sculptured for no less than two aspects (E+S+W for SZX01-08 and ENE+W for BZ4-10-1.2). Based on measurement of ring widths and cross-dating, the periods analyzed were the maximum growth periods; AD1904-1908/1924-1928/1944-1948 for SZX01-08 and AD1930-1944 for BZ4-10-1.2, according to polynomial fitting of ring-width sequences. The holocellulose and α-cellulose fractions were extracted using the Soxhlet method, and the purity of the fractions was confirmed using a NEXUS670 Fourier transform infrared spectrometer. Stable carbon isotope ratios in different components (α-cellulose, holocellulose and wholewood) of samples were measured using a ThermoFinnigan-Delta^plusXP mass spectrometer and expressed as δ¹³C relative to the Vienna Pee Dee Belemnite (VPDB) standard. On the basis of relative analyses, the following results were obtained: (1) The δ¹³C values were highest for α-cellulose followed by holocellulose, and lowest for wholewood. In general, the differences in δ¹³C values among the different components were greater in LW than in EW, as demonstrated by the standard deviations. All differences were statistically significant at the 0.001 level as determined by ANOVA tests. (2) The scatter plots of α-cellulose vs. holocellulose and α-cellulose vs. wholewood showed that holocellulose was more similar than wholewood to α-cellulose, in terms of δ¹³C values. There were no significant correlations between α-cellulose and holocellulose or between holocellulose and wholewood for EW. In contrast, correlations between the δ¹³C values of the three components were all statistically significant at the 0.01 level for LW. This indicated that the isotope signals of LW in tree rings are more coherent and more sensitive to changes in the local climate and environment than EW. (3) To some extent, the climate or environment signals reflected by the δ¹³C ratios are more significant for holocellulose than for α-cellulose. This probably implies that the climate signals were impaired during extraction of α-cellulose. The relationship between δ¹³C of wholewood and climate or environment factors was statistically insignificant. The extraction of α-cellulose is time consuming and results in decreased accuracy; therefore, it is preferable to base palaeoclimatic or palaeoenvironmental reconstructions on data obtained from holocellulose in LW.