Music Between Two Minds: Individual Signatures, Shared States, and the Emergence of Creative Interaction

The gist

Improvising pairs neither fuse into one system nor remain two independent ones, but preserve individual melodic voices while sharing a rhythmic state, within which moment-to-moment push–pull continually rebalances their interaction<ref:2610.31131#pg6>.

How it works

The study examined the information dynamics of interpersonal improvisation using 986 professional jazz improvisations to understand how two minds jointly regulate the flow of novelty in real time<ref:2610.31131#pg3>. The analysis used a general predictive model of jazz expectation, trained on the Weimar Jazz Database, to measure pitch-based and rhythm-based fluctuations for every note in the performance<ref:2610.31131#pg9>.

The researchers found that pitch-based fluctuations behaved as stable signatures of individual identity, meaning the two partners differed within 90% of takes and performer identity could be decoded from single solos <ref:2610.31131#pg1>. In contrast, rhythm-based fluctuations showed the reverse organization, where partners were statistically indistinguishable within a take yet shifted together from take to take, indicating that the rhythm properties belong to the performance itself rather than individual identity.

How it works

The shared rhythmic state is not static but drifts gradually across each performance, with both performers moving together, as the two performers were travelling together, and the shared rhythmic state itself drifted gradually across the take <ref:2610.31131#pg2>. This drift is evidenced by tracking how dissimilarity between solos grew with the number of turns separating them, showing that both series climbed together in rhythm.

How it works

At every solo hand-off, the partners showed a systematic push–pull complementarity, where when one performer exceeded their own average surprise in the predictive dynamics of improvisation, the partner's next solo fell below their own. This coupling was found to be reliable, as real duos showed reliable negative coupling that both controls abolished when compared against null models.

How it works

The central pattern is a double dissociation where melodic fluctuations carried a persistent individual identity: partners never converged, their separation neither narrowed with turns nor shifted with takes while rhythmic fluctuations carried a collective state: indistinguishable between partners, specific to each real interaction, drifting gradually within each performance <ref:2610.31131#pg4>. This organization is described as field intelligence, which is structure in predictive fluctuations that cannot be attributed to either partner alone.

The results suggest that creativity in interaction is sustained by the simultaneous maintenance of individual and shared predictive dynamics carried by different musical dimensions. This division of roles across dimensions allows a duo to be maximally close and reliably distinct at the same time.

The paper concludes that interpersonal improvisation depends not on convergence toward a single common state, but on the simultaneous maintenance of individual and shared predictive dynamics. The organization reveals how novelty is jointly regulated, preserving individual identity while generating shared and novel structure that emerges only at the level of the dyad.

Limitations include the fact that the melodic identity signature may partly reflect instrument-specific affordances rather than personal style alone and that all performances were improvisations on a single tune within a single idiom, which means the findings may not hold across all repertoires, genres, and cultures. The study establishes that the observed coupling exceeds stringent null models, but it does not establish the causal pathway that produces it. The analysis is also restricted to symbolic note sequences, meaning the findings may not hold across all interactions that can be represented as a sequence of events under a shared predictive model. The reference model was trained on a specific historical corpus of monophonic solos, and different reference corpora could shift absolute values of surprise and uncertainty. The results do not yet show that this regulation is what makes the interaction creative in the evaluative sense. The organization revealed suggests that innovation arising from the interplay of shared states and preserved individual variation is key to collective intelligence. The findings predict that some informational dimensions of speech should identify the speaker and resist convergence, others should belong to the particular conversation and drift with it in dialogue. The results show how novelty is jointly regulated, but they do not yet show that this regulation is what makes the interaction creative in the evaluative sense.

--- Page 5 ---

Abstract

Some of the most valued forms of human creativity arise between people—in conversation, collaboration, and improvised music—yet how two minds jointly regulate the flow of novelty in real time remains unknown. Here we examined the information dynamics of interpersonal improvisation using 986 professional jazz improvisations. Pitch-based fluctuations of uncertainty and surprise behaved as signatures of individual identity: the two partners differed within 90% of takes, performers were identifiable from single solos and the separation between partners persisted undiminished across turns and takes. Rhythm-based fluctuations showed the reverse organization: partners were statistically indistinguishable within a take yet shifted together from take to take, and this shared state drifted gradually as each performance progressed, with each performer moving away from their own past at the same rate as from their partner's. Beyond joint drift, solo hand-offs revealed turn-by-turn complementarity: when one performer exceeded their own average surprise in the predictive dynamics of improvisation, the partner's next solo fell below their own, in both pitch and rhythm and in all duos. Improvising pairs thus neither fuse into one system nor remain two independent ones: they preserve individual melodic voices while sharing a rhythmic state, within which moment-to-moment push–pull continually rebalances their interaction.

--- Page 8 ---

Results

To measure the predictive fluctuations that improvisers generate for one another, we first constructed a general model of jazz expectation and then let it listen to real duo interaction, continuing to update its statistics as a listener accumulating experience would. A hierarchical Bayesian statistical-learning model (HBSL; Daikoku et al., 2023) was trained on the Weimar Jazz Database of the Jazzomat project—456 professionally transcribed solos by 78 musicians (Frieler et al., 2013)—separately for three event streams: pitch, rhythm, and their conjunction (Materials and Methods). For every note of a new performance, the model outputs the surprise of the realized event (information content: the negative log probability of the note given its context) and the uncertainty of that context (the entropy of the model's predictive distribution), both in bits.

Pitch-based predictive fluctuations behaved as stable signatures of individual identity, meaning within a single take, the note-level uncertainty values of the two partners differed significantly in 90% of takes. The separation between partners was large and persistent, with distances between solos by the two different performers were 2.14 times greater than distances between solos by the same performer.

Rhythmic fluctuations index which performance is unfolding, as rhythmic fluctuations changed from performance to performance: the same performer differed from themselves across consecutive takes in 60% of comparisons. The take structured both performers together, and only performers who actually played together shared it.

How it works

The study found that melodic fluctuations index who is playing while rhythmic fluctuations answer which performance is unfolding: each take establishes a rhythmic state that both performers jointly occupy—a property of the pair and the occasion that cannot be assigned to either individual.

How it works

Solo hand-offs reveal turn-by-turn complementarity, as in rhythm, the mean within-duo correlation across consecutive solos was r = −0.230 and in pitch, r = −0.151

1. Predictive dynamics between improvising minds can be used to create AI systems that maintain both individual identity and shared context simultaneously. The improved system can generate novel outputs by preserving individual melodic voices while sharing a rhythmic state, within which moment-to-moment push–pull continually rebalances their interaction.

2. The system can dynamically regulate novelty by using the observation that when one performer's surprise was higher than usual, the partner's next solo tended to be less surprising. This allows the AI to manage the flow of unpredictability within a creative generation process, ensuring that deviations are informative yet interpretable.

3. AI models can be structured with two distinct functional dimensions: one dimension for preserving individual identity (like pitch-based fluctuations) and another for generating a shared state (like rhythm-based fluctuations). This allows the AI to simultaneously maintain unique stylistic elements while participating in a collective structure specific to the current interaction.

4. The system can utilize turn-level complementarity by implementing a mechanism where one component's deviation predicts the next component's, as shown by the finding that a performer who played above their own average was reliably followed by a partner playing below their own. This allows for active, predictive coordination rather than simple convergence or independence.

5. AI can implement a field intelligence framework where organization is specific to the particular occasion of playing together, meaning the system's emergent structure is not just about the individuals but about the particular occasion of playing together. This enables the system to create novel, context-specific structures that cannot be reduced to either partner alone.

Some of the most valued forms of human creativity arise between people-in conversation, collaboration, and improvised music-yet how two minds jointly regulate the flow of novelty in real time remains unknown. Here we examined the information dynamics of interpersonal improvisation using 986 professional jazz improvisations. Pitch-based fluctuations of uncertainty and surprise behaved as signatures of individual identity: the two partners differed within 90% of takes, performers were identifiable from single solos and the separation between partners persisted undiminished across turns and takes. Rhythm-based fluctuations showed the reverse organization: partners were statistically indistinguishable within a take yet shifted together from take to take, and this shared state drifted gradually as each performance progressed, with each performer moving away from their own past at the same rate as from their partner's. Beyond joint drift, solo hand-offs revealed turn-by-turn complementarity: when one performer exceeded their own average surprise in the predictive dynamics of improvisation, the partner's next solo fell below their own, in both pitch and rhythm and in all duos. Improvising pairs thus neither fuse into one system nor remain two independent ones: they preserve individual melodic voices while sharing a rhythmic state, within which moment-to-moment push-pull continually rebalances their interaction. Individuality and togetherness thus coexist, carried by different musical dimensions, a measurable organization of creativity that emerges between minds.