Memory can be broadly divided into short and long term. Although properly memorised music is stored in long term memory, it must be retrieved and processed in short term working memory as a piece unfolds in real-time during performance. Working memory allows us to temporarily store and manipulate information, providing the crucial link between transient sensory memory and long term storage. We use working memory when we’re doing mental arithmetic, translating a foreign language, or playing music. Psychological tests of working memory often involve memorising a tray of items or a sequence of words, then recalling them after a short delay. But one could equally ask people to memorise a short tune and sing it back. Would that be any different?
Numerous models have been developed to explain how information is stored and processed in working memory. Perhaps the most influential is the multi-component model developed by cognitive psychologists Baddeley and Hitch , which comprises four separate but interconnected modules:
- a central executive, the master controller which focuses our attention;
- a phonological loop for temporary storage and rehearsal of verbal and auditory information;
- a visuo-spatial sketchpad for temporary storage of visual and spatial information, e.g. objects and their locations; and
- an episodic buffer which allows these different elements to interact with each other and information in long term memory to facilitate comprehension.
The storage capacity in working memory is limited. Only a small number of items can be held in this short term memory at any one time, and we easily forget information either through decay or distraction (e.g. someone shouting numbers at you when you’re tying to remember a phone number), though working memory can be refreshed by mental rehearsal. How is music is processed in the working memory? Does it use the phonological loop – in common with words – or is there another separate loop for tonal memory? This question is particularly relevant in light of recent evolutional theories proposing that the origins of language are partly grounded on cognitive systems dedicated to the processing of pitch . A few different studies have tried to address this vexed issue by comparing verbal and tonal working memory in musicians and non-musicians, using either experimental behavioural psychology or neuroimaging .
The first approach is based around the well proven fact that sequences of rhyming letters or words (B, V, E, G, C) are harder to recall accurately than dissimilar ones (F, K, L, R, X) due to acoustic confusion . Does this phenomenon of phonological similarity have a musical correlate? Williamson et al.  investigated whether short-term musical memory can be similarly confused by asking whether a melody based around tones that are close together is harder to recall than one that jumps around. If so, this might indicate that tonal information is stored in the same phonological loop as words. Subjects were presented with series of either letters or notes of varying degrees of similarity, and the accuracy of their recall tested after a short delay. Although the same acoustic confusion was replicated in sequences of either words or tones in non-musicians, pitch proximity had no effect on recall in musicians. This result was explained by suggesting that although there is substantial overlap between the verbal and tonal working memory system, musical training provides additional (non-auditory) strategies for memorising music.
Independent studies in the world of neuroscience point towards a similar conclusion. Using functional MRI, Schulze et al.  showed that the same core structures of the brain are involved in both verbal and tonal working memory, but additional discrete areas are activated by either verbal or tonal working memory in trained musicians. This suggests the existence of two separate but highly overlapping systems for tonal and verbal memory in musicians, created by musical training. Unsurprisingly, musicians were significantly better at recalling melodies than non-musicians, though both groups performed equally well at verbal recall. The phonological loop is primarily lateralised to the left-hemisphere, whereas the additional brain areas activated in musicians include structures in the right hemisphere involved in control, programming and planning.
These experiments offer tantalising clues about how tonal and speech-based information is temporarily stored and processed in the working memory. But there is still much more to be explored, and some obvious questions arise immediately. If there is a single phonological loop involved in tonal and verbal working memory, how does background music interfere with our ability to read, write or talk? If, on the other hand, there are different loops for verbal and tonal working memory, how do they interact in singers or in highly tonal languages, and can musicians use this independence to their advantage?
 Williamson, V.J., Baddeley, A.D. & Hitch, G.J., Musicians’ and nonmusicians’ short-term memory for verbal and musical sequences: Comparing phonological similarity and pitch proximity, Memory & Cognition 38 (2010); 163-175