VERB AGREEMENT ERRORS IN WRITING

To account for subject-verb agreement errors which can be observed in sentences written by literate adults (e.g., The daughter of the neighb ors come), the present study examined a two-step processing model. The first step consists of automatic activation of a singular or plural v erb form depending on whether the closest preverbal noun is singular o r plural; as the outcome of this first step, the verb agrees with the nearest noun, giving rise to what has been called <<proximity concord> > errors. A second step prevents the occurrence of erroneous subject-v erb agreements that would be unavoidable if the writer had no control on the outcome of the first step. This second step consists of an edit ing-check process allowing the writer to carry out a pre-graphical edi ting of the agreement. Two experiments are reported in which subjects were asked to perform simultaneously two tasks: a) to transcribe a sen tence that was orally presented and, b) to mentally add numbers that t hey heard during the transcription. The results of these experiments s upported two predictions of the processing model of number agreement i n writing: a) the editing-check step consumed cognitive resources; the agreement errors were greatly increased by the temporary cognitive ov erload (as is the case in a double task paradigm) preventing the subje cts from carrying out the control processes; b) the implementation of the editing-check stage depended on the prior detection of potential e rrors, which occurs, for example, when the preverbal item is not a sem antically plausible subject of the verb.