The effect of instructions to use imagery and sentence strategies in verbal discrimination learning (VDL) was investigated in two experiments. In Experiment I, both imaging to and constructing a sentence for the correct item of each pair facilitated VDL relative to an uninstructed control condition. Incorporating both words of the pair into a compound image or compound sentence had no effect, a result that was replicated in Experiment IL Of the two compound instructions. only com pound sentences consistently facilitated associative recall of the VDL pairs, suggesting that compound imagery instructions were not always adhered to by the Ss. Rowe and Paivio (1971a) have provided evidence for the facilitative effect of imagery instructions in verbal discrimination learning (VDL). In that experiment, different groups of Ss learned a list of high-imagery (High-I) words under instructions to either (a) image to only the right (R) word of each pair or (b) form a compound image incorporating both the Rand wrong (W) words, but at the same time "tagging" the image for the R ward as the larger of the two. Performance in the single image condition surpassed that of the compound image group, as weil as an uninstructed control condition and arepetition condition where the Ss overtly rehearsed the Rand W items of each pair in a ratio of 4: I. The compound image condition did not differ from the control in VDL but was the only condition that enhanced performance on a subsequent associative recall test.A reliable effect 01' single image instructions on VDL has been verified in subsequent research (Levin, Ghatala, Wilder, & Inzer, 1973;Rowe, 1972), but the situation with respect to cornpound imagery remains somewhat unclear. As pointed out by Rowe and Paivio (197Ia), the finding that one type of compound image/ tagging strategy is ineffective does not disallow the possibility that some other form of compound imaging might be facilitative. The determination of whether or not this type of strategy can affect VDL is important in view of the striking effects that result from compound imagery in other memory tasks, especially paired associate learning (Paivio. 197L Chapter 10). If compound imagery can be shown to produce different effects in VDL and paired associate learning (PAL), as the initial results suggest. this could help to theoretically extricate the processes involved in discrimination and associative learning 01' pairs. One purpose of the present study was to provide additional evidence on compound imagery effects in VDL through the use of instructions that were nonspecific with regard to ways in which the image corresponding to the R word could be tagged. A