We propose A-Blocks, a novel building block device that enables detection and recognition of children's actions and interactions when building with blocks. Quantitative data received from constructing and breaking A-Blocks can be valuable for various assessment applications (e.g., play therapy, cognitive testing, and education). In our prototype system, each block embeds a wireless measurement device that includes acceleration, angular velocity, and geomagnetic sensors to measure a block's spatial motion and posture during children's play. A standard set of blocks can be managed via Bluetooth in real time. By using combined sensor data, the system can estimate how to stack the blocks on each other by detecting surface collisions (Figure 1) and recognize many fundamental play action patterns (e.g., moving, stacking standing, waving) with SVM. Unlike existing block-shaped devices with physical constraints on their connections (e.g., electrical hooks, magnets), our solid and traditional-shaped block device supports flexible block play that could include more delicate motions reflecting a child's inner state (e.g., learning stages, stress level, representation of an imagination). These benefits of analyzing children's block play can be extended to allow for more enjoyable and interactive play, while social impacts include more constructive play.