Abstract. Computer Assisted Transcription of Text Images (CATTI)and Key-Word Spotting (KWS) applications aim at transcribing and indexing handwritten documents respectively. They both are approached by means of Word Graphs (WG) obtained using segmentation-free handwritten text recognition technology based on N -gram Language Models and Hidden Markov Models. A large WG contains most of the relevant information of the original text (line) image needed for CATTI and KWS but, if it is too large, the computational cost of generating and using it can become unaffordable. Conversely, if it is too small, relevant information may be lost, leading to a reduction of CATTI/KWS in performance accuracy. We study the trade-off between WG size and CATTI & KWS performance in terms of effectiveness and efficiency. Results show that small, computationally cheap WGs can be used without loosing the excellent CATTI/KWS performance achieved with huge WGs.