Work sampling has been accepted as a valuable method for assessing the productivity of workforce in construction sites. It helps in identifying the problem areas and gives direction for productivity improvement programmes. But, observing and categorizing work activities is a tedious and difficult task in work sampling assignment. Video-based methods have been investigated as an alternative for human observation, but their performance in classifying worker activities is severely affected by the harsh construction site environment. Preliminary studies have shown that accelerometer data carry rich information pertaining to movement characteristics of workers and hence can be used classifying worker activities. The present study is an extension to this, investigating accelerometer based method for classifying bricklaying activities. The data collection was performed on five bricklayers with tri-axial accelerometer data loggers attached to the right left lower arm, the left lower arm and the waist, during their normal course of work. Video recordings of the worker activities were simultaneously carried out to serve as the ground truth. The accelerometer data was subjected to preprocessing and also trained with multilayer perceptron classifier algorithm for worker activity classification. The results show that the accelerometer-based classifier gave an overall accuracy of 81.37% while classifying bricklaying activities into classes of effective, contributory and ineffective categories. © 2012 ASCE.