Poisonous plants and noxious weeds are often chemically examined to determine concentrations of secondary metabolites which are responsible for their toxic or biological activity. This study examined sample size requirements and sample methods necessary to quantify accurately the concentrations of individual and total toxic alkaloids in two tall larkspur populations. A high performance liquid chromatography analytical method was utilized to determine toxic alkaloid concentrations in all leaves from three individual plant stems and leaves from the remaining stems (remainder) from each of 50 plants in each population. To obtain high precision in quantifying toxic alkaloids in the larkspur populations (within 2.5 to 5% of the population mean, 0.95 confidence), very large numbers of samples (>50–200) were required. However, lower precision (within 10% of the population mean, 0.90–0.95 confidence) required only 20 samples. Similarly, testing parameters relating to toxin concentration in tall larkspur populations within 5 or 10% of the population mean also required hundreds of samples at power levels of 0.95 and α-levels of 0.05. Relaxing power and α-level requirements to 0.80 and 0.1 respectively, reduced sample size to about 30. The means obtained by four different sampling methods were similar (P>0.05). Alkaloid concentrations in leaf samples from single stems were highly correlated to whole-plant leaf (remainder) samples (r2≥0.76), indicating that harvesting leaves from single stems provided representative samples of the entire plant. The results indicate the difficulty in obtaining accurate information about toxins in poisonous plant populations for risk assessment by livestock producers or extension agents and demonstrate the necessity for efficient analytical methodology. Researchers evaluating concentrations of plant compounds in other weeds or toxic plants should consider variability, sampling procedure, and sample size before experiments begin.