Barbara Club
Together with Zofia Holubowska and Lea Fink, we organised the “Barbara Club.” Fueled by coffee, cake, and good vibes, we read and discussed several articles of Barbara Tillmann’s group, sharing thoughts on music structure, perception, phrase tracking, and related topics. The conversations were lively and deeply enriching, informing and inspiring our own ongoing research.
Credit to Zofia for the notes of our discussions that are summarized below.
Warning: These summaries are not exhaustive and do not aim to reflect Barbara Tillmann’s own views, but rather our humble interpretation of a selected subset of papers, shaped by our interests and discussions.
Music unfolds through alternating patterns of tension and resolution, with cadences playing a central role in marking phrase endings. While listeners are known to be sensitive to the degree of closure/completeness conveyed by different cadences, this study addressed a more specific question: how much context do listeners use when forming these judgments? To investigate this, Tillmann and colleagues introduced an original musical puzzle paradigm. Sixteen-bar excerpts from minuets by Bach, Haydn, and Mozart were divided into two 8-bar halves. The first half ended either with a half cadence (locally incomplete) or with an authentic cadence in the dominant key (locally complete but globally unresolved). Participants were asked to match the first halves with their correct second halves. The results showed that matching accuracy was higher when the first half ended with a half cadence, suggesting that listeners treated these endings as incomplete and expected continuation. In contrast, authentic cadences in the dominant key were often perceived as complete despite their unresolved global function. These findings indicate that local harmonic function can outweigh global structural interpretation. Besides clarifying an important point, this study describes an interesting method using musical puzzles to investigate participants’ decision process (instead of the usual evaluation at the end of the presented material), which inspired numerous researchers (e.g., Granot & Jacoby, 2012; Erel, 2022).
Building on the question of how context shapes perception, this study shifted from cadences to target detection, in the vein of studies in the visual domain (see Biederman et al., 1973) where the task is to detect the target in an incongruent environment. The authors asked whether disrupting global musical structure affects listeners’ ability to recognize a previously learned melodic fragment. Sixteen-bar minuet excerpts were presented either in their original form or in altered versions in which fragments were shuffled, sometimes with additional transpositions to increase structural incoherence. Participants were asked to detect a target melody learned beforehand. Results showed that detection was more difficult in structurally incoherent contexts. This is in line with the work of Mandler (1980) on memory model, with the existence of two processes: (1) fast testing of familiarity value and (2) slower testing and detecting if a target was presented before. Presumably, in the case of music, the first process seems to be particularly important and recognition does not appear as linked to the global structure of the piece. As a follow up to this study we may wonder what the results would be if participants would not be asked the explicit target-detection question “Have you heard that before?”, but using EEG to capture the perception of something know (if possible). Also, it would be certainly interesting to manipulate the coherence of the environment as well as the strength of the alteration in order to generalize such findings to a larger variety of musical material.
While the two other studies focused on completion and recognition, this article directly addressed expectations using a priming paradigm. Participants listened to eight-chord sequences and judged whether the final chord sounded consonant or dissonant. The harmonic context preceding the target chord was manipulated either globally (the entire sequence) or locally (the two chords before the target). As expected, judgments were most accurate when both local and global contexts were congruent with the target chord. Importantly, expectancy effects emerged even with minimal context: very short local contexts were sufficient to shape listeners’ expectations, echoing the role of local processing observed in the earlier puzzle and target-detection studies. At the same time, global harmonic context continued to influence perception even when local violations occurred, suggesting a more complex interaction between levels of structure. A potential follow up (just a thought.. perhaps already done) could be to examine the role of music preference of the participants and music style (beyond the usual classical music) to generalize such findings.
This study returned to the harmonic focus of previous work with a more fine-grained perceptual task: detecting a mistuned octave in the final chord of an eight-chord sequence. Two factors were manipulated: the order of the chords (original vs. scrambled) and the harmonic relatedness between the first six chords and the final two. Scrambling the chord order reduced perceived coherence but did not affect mistuning detection. In contrast, harmonic relatedness had a strong effect: mistuned chords were detected more easily when the harmonic context fulfilled listeners’ expectations. This pattern reinforces earlier findings by showing that harmonic context, rather than surface order or coherence, drives perceptual sensitivity. The fact that similar effects were observed in musicians and non-musicians further supports the idea (already implicit in the 1998 studies) that harmonic expectations rely on broadly shared, implicit knowledge rather than formal training.
This article provides a theoretical synthesis of the issues raised in the preceding empirical work. Tillmann and Bigand address two recurring paradoxes: why musical expertise does not strongly enhance processing efficiency, and why listeners appear to process short musical units more precisely than large-scale structures. The authors argue that listeners possess robust implicit knowledge of Western tonal regularities, enabling rapid formation of expectations even without formal training. Across tasks, local context consistently dominates, while global structure is represented more approximately. Experimental evidence suggests that information from short musical excerpts is not necessarily integrated into a unified large-scale representation. This perspective offers a unifying framework for the earlier findings: music perception appears to operate primarily within short temporal windows, with local harmonic information playing a central role in shaping expectations and perceptual judgments.
To conclude..
Taken together, these articles (among others) set solid ground in the study of musical expectations. Beginning with innovative paradigms that dissociate local and global harmonic functions, such work demonstrates that local context plays a decisive role across a wide range of perceptual tasks, including completion judgments, target detection, consonance evaluation, and mistuning perception, while still acknowledging the relevance of global structure. This group of smart researchers have clearly inspired other studies, including our own work on phrase tracking in both regular (Teng, Larrouy-Maestri, & Poeppel, 2024) and irregular musical contexts (Holubowska, Teng, & Larrouy-Maestri, 2025, preprint).
Last but not least, revisiting these papers together was genuinely fun and it definitely makes me look forward to the next XXX (you?) Club!