• Eliminating Two Sources of Heterogeneity in Experimental Design:
• Latin Square Design (LSD) arranges treatments in an $m \times m$ square such that each treatment appears once in each row and once in each column.
• This simultaneously eliminates two orthogonal sources of variation (row variation and column variation, such as fertility gradients in two directions).
• Other Designs:
• CRD: Zero sources of restriction.
• RBD: Eliminates only one source of variation (blocks).
• Conclusion: LSD only (Option 4).

Correct option: (4) LSD only