The fact that inter-observer reliability for the transcripts and data were calculated for only 10% of the data is another limitation; typically, at least 20% of data involving child transcripts are used for inter-observer reliability measures (e.g., Fey et al., 1993). In addition, coders were not blind to the hypotheses or phases of the study, which may have biased the findings. Also, one instructor (i.e., the first author) provided instruction with all partici- pants, and the intervention took place within limited contexts, thus restricting the external validity of the investigation. Furthermore, it is not known if the children would have generalized use of symbol combinations to other contexts (e.g., activities of daily living).