Digital Mind 🧠

🪵 The Threshold

Auditory change detection integrates statistical evidence accumulated over time

Properties2
tagsja-picks, xref, define, seed, question, causal_relationships, trade, corroborated, category/causal-relationships
aliases@van_dijk_physiology_2016, loc. 3285

5 min read

Auditory change detection integrates statistical evidence accumulated over time

Evidence

The above results are consistent with predictions from statistical integration of information. First, unsurprisingly, change detection should improve with greater change sizes. Second, change detection should improve, but saturate, as the estimation of the statistics of the first stimulus converges, i.e. if longer observation time is available. Third, change detection of bigger changes should require less precision in the estimation, translating to higher hit rate already for shorter observation times. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3285

In accordance with the perceived difficulty of the detection task, d’ values are significantly positive (Fig. 1h , based on the 95 % confidence bounds) and increase with change time (Friedman test, p < 0.01), but stay relatively low especially for 110 % and below, indicates how challenging the task was, leading to a substantial number of false alarms. The above results are consistent with predictions from statistical integration of information. First, unsurprisingly, change detection should improve with greater change sizes. Second, change detection should improve, but saturate, as the estimation of the statistics of the first stimulus converges, i.e. if longer observation time is available. Third, change detection of bigger changes should require less precision in the estimation, translating to higher hit rate already for shorter observation times. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3327

The size of change as well as the time available to sample the original statistics were found to correlate positively with performance and negatively with reaction time, suggesting the accumulation of noisy evidence. In summary we quantified dynamic aspects of change detection in statistically defined contexts, and found evidence of integration of statistical information. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3280

We hypothesized that reaction times here could be informative. Reaction times had previously been demonstrated to depend on task difficulty (Kiani et al. 2014 ), and difficulty in the present paradigm correlates inversely with the availability of a greater difference in change size, i.e. statistics (Fig. 2a ). Fig. 2 Reaction times reflect task difficulty and temporal integration. a –c If reaction times are separated by change size (a ), one observes a progression in the shape (b ) as well as median response time (c ). The shape of the reaction time distribution sharpens with increasing step-size and the median response time decreases significantly (p = 1.51e-05, Kruskal-Wallis). d –f If reaction times are separated by the duration of the change time before the change in statistics occurred (d ), the distribution of reaction times again sharpens with longer change times (e ) and the median reaction time (f ) decreases significantly (p = 5.46e-06, Kruskal-Wallis) Reaction time distributions changed both in time and shape as a function of change size (Fig. 2b ). Median reaction time correlated negatively with change sizes (p < 0.001; Fig. 2c ), in a manner that corresponded to the inverse of the increase in hit rate with larger change sizes. In addition, the onset time of the distributions decreased with change size, together suggesting that certainty was reached earlier for bigger step-sizes. As a function of change time, the reaction time distribution changed in a qualitatively similar manner, however, less pronounced (Fig. 2e ). Median reaction times correlated negatively with change times, mirroring dependence of hit rate on change times (p < 0.001; Fig. 2f ). This dependence can already be seen in the raw data (Fig. 1c ), where hit trials (black) for longer change times exhibit shorter reaction times. Again the onset time decreased correspondingly with longer change time, suggesting more accurate estimation of the initial statistics. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3332

Categorization

  • XREF: Parallels sequential sampling and drift-diffusion models in decision making; also bounded accumulation models from cognitive psychology. — connects to: Statistically-driven change detection may improve when attention is withdrawn.; Detecting changes in auditory textures depends mainly on integration time available; Detection of statistical change requires temporal integration of the marginal distribution (+10 more)
  • DEFINE: Explains the ‘statistical integration’ framework: detection improves with observation time but saturates as estimation converges, and bigger changes need less estimation precision.
  • SEED: A behavioral result on accumulation of noisy perceptual evidence could anchor an essay on how perception trades sampling duration against change magnitude.
  • QUESTION: Reaction time is a proxy for when the decision system reaches internal certainty — how early decision onsets map onto accumulation thresholds in drift-diffusion terms is worth chasing.
  • corroborated: 7 independent passages support this claim — Dijk, Pim van, Physiology, Psychoacoustics and…, 3285; Dijk, Pim van, Physiology, Psychoacoustics and…, 3327; Dijk, Pim van, Physiology, Psychoacoustics and…, 3280 (+1 more)

Broader Connections

✉️

Backlinks