Recent results with memory for real pieces of music have shown improvement in recognition performance over delays of 3-15 s after the initial presentation of a musical phrase (Dowling & Tillmann, 2014; Dowling, Tillmann, & Ayers, 2001; Tillmann et al., 2013). In a typical study, listeners heard the beginning of a classical minuet. Toward the start of the minuet, there was a target phrase that would be tested later. The minuet continued just as the composer had written it, without repeating or imitating the target phrase. After a shorter or longer delay filled with the continuation of the minuet, the listener heard a high-pitched beep that signaled the onset of the test. The test phrase that followed was the same as the target phrase (T), a similar lure (S) sharing the melodic-rhythmic contour of the target, but differing from it in detail (typically a shift in pitch of the melodic line), or a different lure (D) having a different melodic-rhythmic contour. Listeners had to decide whether the test phrase was exactly the same as an earlier phrase they had heard in that piece, or different. Discrimination of targets from similar lures (T/S discrimination) improved with delay between 3 s and 15 s. This improvement was principally due to a decrease in false alarms to similar lures while hit rates remained constant over time. That is, with increasing delay, listeners improved their discrimination of targets and similar lures. The first three experiments shown in Table 1 provide some typical results from those experiments.We attributed this pattern of responses to the continued encoding of the musical phrases while listeners continued to follow the piece, binding together separate features encoded individually when the listeners had first heard the target phrase (Dowling & Tillmann, 2014; Tillmann & Dowling, 2007; Tillmann et al., 2013). For example, the melodic contour and the musical scale in a particular key would be encoded separately as individual features early in the memorization process.1 When a decision concerning an S lure test item is made at the shorter delay, the memory system would rely on the considerable overlap of individual features between T and S, and accept the similar lure as the same as the target (leading to higher false-alarm rates). After a longer delay, during which additional encoding of the target phrase (and notably, the binding together of the features) continues, the contour and the scale would become bound together. This binding would produce a coherent memory representation of the phrase that could be stored as a unit in memory, retrieved when needed, and manipulated in working memory. Because of the binding of the individual features, differences between the S lure and the target would become apparent and the S lure would be rejected (leading to a decreased rate of false alarms). An example of the contour of a target phrase is shown in Figure 1. The contour of an S lure is labeled S. The pitches of this contour have been shifted up or down, while the contour has maintained the same shape. When tested at the short delay, the individual features of contour and key are sufficient for the memory system to return a positive response in comparing the S lure with the target (i.e., a false alarm). However, after a longer delay, and thus after additional encoding in which the contour is bound to the scale at the appropriate pitch in the representation of the target, the S lure is easier to reject and a false alarm is more easily avoided. It is only when the contour and tonal scale are combined, with the contour being bound to the scale at a particular pitch level, that the particular musical structure underlying the phrase is clarified (Dowling, 1978). In summary, when memory is tested before the contour and scale features have been bound together, there is confusion between Ts and Ss, as they share a high proportion of individual features. Later, when the features have been bound, the differences between the target and similar lure phrases become apparent. …
Read full abstract