From 4 notes to 16: a working-memory approach to melodic dictation
Why most learners hit a wall in melodic dictation around six or seven notes — and how chunking, working-memory limits, and the right scaffolding produce a way through.
Most learners hit a wall in melodic dictation in roughly the same place. Phrases of three or four notes are manageable. Phrases of six or seven start to slip. By the time a phrase is twelve or sixteen notes, the entire melody seems to evaporate from memory before the pen reaches the paper. The frustrating part is that the learner can usually recognize the phrase if they hear it again. They simply can’t hold it long enough to write it down.
This pattern is not a personal failing. It is exactly what cognitive psychology predicts. Understanding it is the first step toward training around it.
The working-memory ceiling
Working memory — the system that holds and manipulates information across short timescales — has a famously limited capacity. Miller’s “magical number seven, plus or minus two” (1956) was the original estimate; later work has revised this downward, with Cowan’s review converging on roughly four chunks for unstructured items [1]. Music is a streaming auditory signal that arrives faster than a learner can store it as discrete pitches. Without intervention, melodic dictation runs head-first into this ceiling.
A 2025 Frontiers in Psychology study by Baker and colleagues confirmed the obvious-but-important: working memory capacity, alongside aural skills training and prior piano experience, was a significant predictor of melodic dictation performance among university music majors [2]. Bigger working memory, better dictation. But this is not destiny. The same study and a broader literature point to a specific cognitive operation that distinguishes learners who break through the ceiling from those who don’t: chunking.
Chunking is what trained musicians actually do
A chunk is a meaningful unit. To a non-musician, C-D-E-F-G is five separate notes. To a trained musician, it is “the first five notes of a major scale” — one chunk. The same five notes, played with the third raised, become “C major to E minor arpeggio outline” — two chunks. By replacing low-level pitch sequences with higher-level musical concepts, the trained musician spends less working memory storing the phrase and frees capacity for further notes.
David Baker’s dissertation work on melodic dictation, which formalized the cognitive model behind this skill, identifies chunking as one of four core sub-skills (alongside working memory itself, transcription speed, and selective attention) [3]. Crucially, chunking is trainable. It is not a fixed property of the learner. It is the deliberate construction, through practice, of a vocabulary of musical patterns that the ear recognizes faster than it processes individual notes.
The progression that breaks the ceiling
A working-memory-aware approach to melodic dictation training looks roughly like this:
Stage 0: Contour only (1-3 notes). Before any pitch identification, train the ear to track direction. Did the line go up, down, or stay? Infants and adults track contour as the most basic feature of a melody [4]. Two- and three-note contour patterns are the lowest-load possible exercise and they install the habit of listening for shape first.
Stage 1: Short phrases (4-5 notes). Once contour is fluent, add pitch. Four notes is comfortably within Cowan’s working-memory bound. Constrain the material — a single key, mostly stepwise motion, no rhythm to track yet. The goal is reliable accuracy at this length, not speed.
Stage 2: Build chunk vocabulary (5-8 notes). Introduce phrases that contain recurring musical idioms — scale fragments, arpeggios, tonic-and-dominant outlines, common cadential figures. Name them out loud as you hear them. (“That’s 1-2-3-2-1.” “That’s an arpeggio of V.”) This is the deliberate chunking step. Each idiom you can name fluently shrinks from N notes to one chunk in working memory.
Stage 3: Compound phrases (8-12 notes). Phrases now combine two or three idioms. The challenge is no longer holding individual pitches — it is holding the sequence of chunks. This is where a learner’s chunk vocabulary from Stage 2 starts to pay off. A phrase that was twelve undifferentiated notes becomes “scale fragment, then arpeggio, then turn.” Three chunks, easy to retain.
Stage 4: Real melodies (12-16+ notes). Now you can take on actual repertoire. By this point the working memory has more leverage on what it stores; the bottleneck shifts from capacity to chunk vocabulary breadth.
This progression takes time. In an undergraduate aural-skills program it spans semesters. The point is not to rush — it is to recognize at which stage you are stuck and direct practice there.
Tactical advice that follows from the science
A few practical implications of the research:
Listen multiple times, but with different goals each pass. Treoria and similar dictation tools allow unlimited replays. Use the first pass for contour (“up, up, down, down, up”). Use the second for cadential structure (“ends on 1”). Use the third for specific scale degrees. This is deliberate sub-task decomposition — a known technique for managing working-memory load [3:1].
Notate as you listen, not after. Hold-it-all-then-write is the maximum-load strategy. Real transcribers write during the listening, even if it’s just contour arrows on a first pass.
Sing what you just heard before you write. Vocalizing serves two functions: it engages the sensorimotor pitch-production system (see our companion article on singing while you train), and it converts a passive memory into an active one, which is more durable.
Pre-establish the key. A cadence or scale before the dictation phrase reduces the load of “figure out the key” so that working memory can be spent on “figure out the notes.” See our article on cadence-first practice.
Practice short, often. The spacing effect (see our article on spaced repetition) applies to dictation as much as to anything else, and short sessions stay within the deliberate-practice attention window where new chunks actually consolidate [5].
What ear-training apps can do better
Most current ear-training apps treat melodic dictation as a single category — drop the learner straight into 8–12 note phrases. Given the working-memory literature, this is poor pedagogy. A staged approach — contour-only → 4-note → chunked-idioms → compound — would mirror what undergraduate programs actually do, and would give learners reliable success at each level rather than a frustrating wall.
Fifths’ Melodic Dictation segment is one of the planned-but-unbuilt segments at the time of writing. We’ve designed it around this staging — contour first, chunks second, compounds third — because the alternative is shipping the same wall of note-by-note transcription that drives most learners off the practice.
The takeaway
Melodic dictation feels impossible at first not because the learner has a bad ear, but because the brain’s working memory is too small to hold raw pitches the way the task seems to demand. The trained musician isn’t using more working memory; they’re using better chunks. Chunks are built deliberately, over time, from short phrases of recognizable musical idioms. If you find yourself stuck at five or six notes, stop trying to push through to eight. Drop back to four, build chunks at four and five, and the ceiling will rise on its own.
References
Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24, 87–185. https://doi.org/10.1017/S0140525X01003922. The original Miller paper: Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. ↩︎
Baker, D. J., et al. (2025). Piano history, aural skills, and working memory predict melodic dictation performance. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2025.1579714 ↩︎
Baker, D. J. (2018). Modeling Melodic Dictation [Doctoral dissertation]. Available online: https://davidjohnbaker1.github.io/document/intro.html and https://davidjohnbaker1.github.io/document/individual-differences.html ↩︎ ↩︎
Dowling, W. J. (1978). Scale and contour: Two components of a theory of memory for melodies. Psychological Review, 85(4), 341–354. See also: Dowling, W. J., et al. (1999). Melodic and rhythmic contour in perception and memory. https://www.decisionneurosciencelab.org/pdfs/Dowling et al., (1999).pdf. On infant contour perception: Trehub, S. E., Bull, D., & Thorpe, L. A. (1984). Infants’ perception of melodies: The role of melodic contour. Child Development, 55(3), 821–830. ↩︎
Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406. Replication and revisit: Macnamara, B. N., & Maitra, M. (2019). The role of deliberate practice in expert performance. Royal Society Open Science, 6, 190327. https://royalsocietypublishing.org/doi/10.1098/rsos.190327 ↩︎