This study investigates the classifier constructions in sign languages with a focus on the properties of the roots. Based on the observations in Turkish Sign Language (TİD), this study claims that roots in classifier constructions do not bear any syntactic, semantic, or phonological information. Merging with a movement morpheme that reflects the actual world movement of an entity, the root gains its meaning through the arguments introduced via the functional heads. This analysis aims to utilize no phenomenon-specific linguistic elements while capturing the characteristics of classifier constructions.