For many years, breeding of Pinus radiata for structural wood relied on improving basic density assessed at age 7 or 8 years old, with little progress. Current efforts have moved to acoustic screening for stiffness at similar age. Breeding cycles are still too long. An alternative is to screen out the worst trees even earlier: shorter breeding cycles should outweigh the lower accuracy due to early selection. Besides genetic effects, there is also evidence that wood stiffness is affected by wind, particularly for stands with low stocking and trees in forest margins. A glasshouse experiment was setup for early selection considering two factors: tree position and clone. Tree positions were straight (control), leaning (30° from the vertical) and rocked (15 minutes every hour, simulating 10 km h-1 wind). Four clones were used covering a range of wood stiffness and replication was 12 plants per treatment. The response variables at 8 months were squared acoustic velocity (v2, surrogate of stiffness), basic density, collar diameter, diameter asymmetry and compression wood. There were significant differences of v2 for treatments and clones. Straight trees had the higher v2 (2.15 km2 s-2), followed by leaning trees (1.95 km2 s-2) and rocked trees (1.74 km2 s-2). The 19% v2 reduction from straight to rocking trees is consistent with observations on the effect of forest margins. Clonal means ranged from 1.53 to 2.11 km2 s-2. Basic density showed significant differences between treatments but not for clones, with higher values for leaning trees (408.0 kg m-3), followed by rocked trees (370.2 kg m-3) and straight trees (358.3 kg m-3). There was zero correlation between v2 and basic density. Straight and rocked trees formed little compression wood in thin arcs at random within the cross-section. Leaning trees formed continuous compression wood on the underside of the leaning stem. We discuss the implications for tree improvement.