Douglas-fir (Pseudotsuga menziesii var. menziesii (Mirb.) Franco) branch segments were used to test the hypothesis that compression wood reduces xylem transport efficiency. Whole 3-year-old segments were first measured for specific conductivity (k(s), m(2) s(-1) MPa(-1)), then split lengthwise into upper and lower halves, the latter containing all or most of the compression wood in the segment. Halves were then remeasured for k(s) using a new technique that prevents leakage of permeating fluid during measurements. Lower branch halves had significantly lower k(s) than upper halves (6.4 +/- 0.3 versus 9.3 +/- 0.3 m(2) s(-1) MPa(-1) x 10(-4), respectively; n = 36), and despite their larger size, significantly lower hydraulic conductivity (k(h), m(4) s(-1) MPa(-1)) than upper halves. Lower branch halves had higher specific gravity (0.51 +/- 0.01 versus 0.45 +/- 0.01; n = 36), lower water content (123 +/- 2% versus 155 +/- 3%; n = 36), and larger proportions of volume occupied by both cell wall and air than upper halves. Lower halves had more tracheids per annual ring than upper halves (73 +/- 3 versus 63 +/- 2 per radial transect, respectively; n = 36), but tracheids were shorter and had narrower lumens than those of upper branch halves. Differences in hydraulic properties between upper and lower halves suggest that compression wood does reduce xylem transport efficiency. In contrast, the amount of compression wood in each sample did not explain any variation in whole unsplit sample hydraulic properties.