According to the ABNT NBR 7190: 1997 standard, tensile strength of wood samples shall be defined as the ratio between the ultimate strength from a stress test and the specimen cross-sectional area (350 mm 2). However, due to the complex wood anatomy and experimental difficulties, the fractured surface in tested samples are not perpendicular to the loading direction; therefore, such fractured surface becomes larger than the sample cross section, overestimating tensile strength by this method. Using an alternative approach, we aimed to determine the approximate fracture surface of wood specimens subjected to tensile test. For this purpose, we used the least square method to estimate tensile strength of four hardwood species, which were equally divided into strength classes. Then, we determined the relationship between the cross-sectional areas of intact and fractured specimens. The results showed that the approximate area of a fractured surface was 2.14 higher than that of an intact sample. As a result, tensile strength estimates were 47% lower than those currently estimated, which is unfavorable for the safety of construction structures.