Different devices have been developed to enable persons with cervical spinal cord injury (SCI) to use a computer. However, text input speed (TIS) remains low for these persons. Several methods have been developed to increase TIS, such as word prediction software (WPS). Data in the literature are discordant regarding the effect of WPS on TIS, with decreases of up to 71% in some studies and increases of up to 45% in others. The main reason suggested for these differences is the cognitive load caused by the visual searching for words in the prediction list. It is thus likely that the number of words in the prediction list influences TIS. Moreover, WPS can be customized using different settings. For example, the prediction list can be ordered according to the frequency of occurrence of words in the language (frequency of use) and the automatic learning of new words can be activated. Until now, no studies in the literature have evaluated the influence of these parameters on TIS. The aims of this study are to determine if the number of words displayed in the WPS list and activation of the frequency of use and automatic learning parameters of WPS affects TIS in people with tetraplegia.

Forty-five persons with cervical SCI between C4 and C8 Asia AIS A or B and who were computer users were included. For the parameter “Number of Words displayed”, TIS was evaluated during 4 10-minute copying tasks (without WPS, with a display of 3 6 8 predicted words). For the parameter “frequency of use”, TIS was evaluated during 3 copying tasks (without WPS (WITHOUT), with automatic learning of words and frequency of use desactivated (NOT_ACTIV) or activated (ACTIV)).

There was no effect of the number of words displayed in a word prediction list on TIS, however perception of TIS differed according to lesion level. Use of word prediction software with activation of frequency of use and automatic learning increased TIS in participants with high-level tetraplegia.

For participants with low-level tetraplegia, use of word prediction software with frequency of use and automatic learning activated only decreased the number of errors.