Christopher E. Siddon and Jon D. Witman. 2004. Behavioral indirect interactions: multiple predator effects and prey switching in the rocky subtidal. Ecology 85:2938–2945.

Appendix B. A description of the cage design and artifact testing.

Five-sided, bottomless cages were constructed to allow consumer species unencumbered access to a 1.0 m² area of substratum. The cages consisted of a 1 × 1 × 0.5 m (L × W × H) base onto which a heavy flexible skirt was attached to fit the irregular contours of the bedrock substrate. The base was constructed from vinyl-coated wire mesh (Aquamesh) with 25.8 cm² (2 × 2 in.) openings. Five centimeters on one end of each side and the two opposite ends of the top were bent 90° to increase rigidity. All four sides were connected to one another using aluminum "J-clips" (Stanley brand). The top was secured with "J-clips" on one end (acting as a hinge) and two pieces of copper wire on the opposite end to allow easy access to census prey and assess habitat changes in the cages.

The skirt consisted of 0.95 cm (3/8 in.) galvanized chain inserted into an equal length of mesh bait bag material and attached along the perimeter of the base using plastic wire ties. The flexible bait bag material allowed the chain to lie flat on the substrate while providing a barrier between the substrate and the bottom of the cage, thus minimizing the escape of experimental organisms.

We examined potential cage artifacts on water flow and irradiance at Smith's Cove. Water flow was measured inside and outside of 10 cages for 30 seconds at 10 Hz using a Sontek 10-MHz Acoustic Doppler Velocimeter (ADV). Measurements were taken during the period of highest flow approximately halfway between high and low tides. Irradiance, measured as PAR, was quantified inside and outside of 5 cages for 60 seconds each using an Integrating Natural Flurometer (INF, Biospherical Instruments). Measurements were taken around noon on a clear day to minimize light variation. Data were analyzed using t tests on the averaged values from each sample. All data met the assumptions of normality and homoscedasticity. There was no significant difference in water velocities (P = 0.667, df = 18) or light levels (P =  0.052, df = 8), however PAR was roughly 25 % lower inside the cages (Table B1). All cages were cleaned manually in situ> as needed to minimize potential artifacts due to fouling over time.

TABLE B1.  Test of cage effects on water flow and irradiance. Values are means (± SE).  N = 10 for water velocity and n = 5 for PAR.

Cage	Water velocity (cm/s)	PAR
Inside	2.98 ± 0.80	24.00 ± 4.34
Outside	3.16 ± 0.31	32.33 ± 6.93