This study investigated (1) whether symbol-symbol correspondences are learned better when imbedded in whole words or when presented individually and (2) the effect of these two instructional methods on transfer from an experimental transitional writing system (TWS) to traditional orthography (TO). Thirty-five undergraduates were randomly assigned to two treatment groups (18 to individual symbol method of presentation and 17 to the embedded symbol whole word method). For the first group, 17 TWS graphemes were paired with corresponding TO graphemes in the Roman alphabet for learning; for the second group, the symbols were arranged to spell nine words. In a transfer test the subjects were required to transliterate seven novel words into their TO equivalents. The number of trials to reach criterion (one errorless trial) for both methods was analyzed by means of ANOVA, and the Mann-Whitney U Test was used to analyze the data from the transfer task. Results showed that instruction by the whole word method resulted in faster learning, but this mastery was not as conducive to transfer as learning the symbols individually. When learning and transliteration were considered together, the individual letter method of instruction was significantly faster than the whole word method. It was concluded that though slower in initial learning, the individual letter method was superior and faster for transfer performance and overall time. Tables and references are included. (AW)