Experimental environment.

A. Speech-in-noise immersed in real reverberant environments reproduced in an anechoic chamber over a 41-loudspeaker array. All speech material was rendered as if it was originating from directly in front of a listener and was masked by spatially separated (90° to the left) white noise at −15 dB signal-to-noise ratio (SNR). The noise masker level was fixed at 70 dB SPL. The speech material was convolved with the measured impulse response of a variety of different real rooms, from anechoic to highly reverberant real rooms such as an underground car park (see Table 1 methods for acoustic details). B. Structure of the speech material. Sentences from the Coordinated Response Measures (CRM) corpus were modified to vary their duration depending on how many words preceded the target phrase i.e., ‘carrier phrase’ (CP) length. Participants were asked to recognize and repeat target words: |Color| from a list of four and |Number| from a 1-8 list. C. Example of the experimental paradigm. Trials consisted of ‘carrier phrases’ spoken in a specific room i.e., ‘go to blue 1 now’ in the Lecture Room. Participants were never exposed to the same room consecutively and the task lasted no longer than 45 minutes. D. Sensitivity to |Color| and |Number| combined in the Lecture Room, Open-Plan Office and Underground Car Park. Mean d’ (measure of accuracy calculated as: Z (correct responses) – Z (false alarm)] denoted as circles in the boxplot (n=22 for all rooms). The horizontal line denotes the median. Upper and lower limits of the boxplot represent 1st (q1) and 3rd (q3) quartiles respectively, while whiskers denote the interquartile range (IQR =q3-q1). As previously reported (Brandewie & Zahorik, 2010, 2013), we observed an increase in performance for longer ‘carrier phrases’ in all environments. https://doi.org/10.25949/24295342.v1. All images in this figure were generated using artificial intelligence, with the exception of the Anechoic Room photograph, which was taken by the authors

Real rooms acoustic characteristics.

Learning effects in three real rooms.

A. Overall Performance (d’) (including correct and incorrect responses to all |Colors| and |Numbers|) in the Lecture Room (LR), Open-Plan Office (OPO) and Car Park (CP). Mean d’ denoted as circles in the boxplot (n=22 for all rooms). The horizontal line denotes the median. Upper and lower limits of the boxplot represent 1st (q1) and 3rd (q3) quartiles respectively, while whiskers denote the interquartile range (IQR =q3-q1). B. Time course of performance in each room for 22 listeners. Correct responses (Hit Rate) across time spent in each room are shown, solid curves in color correspond to mean data after 5-time point moving averages, equivalent to ∼7s, were applied). Optimal time course-fittings (using a two-phase association model) are plotted as color markers in each room. Associated shaded areas representing standard errors of the mean. Dashed coloured lines represent the time point at which participants reached a stable performance in each environment [i.e., ± 10% of the Final Hit Rate (FHR)], here ‘global learning’. Notice that no statistical differences were found for ‘global learning’ among the different rooms. C. Shows a correlation analysis between FHR and the time where 10% of the FHR is achieved in each environment. Significant negative correlations (Pearson) were observed for Lecture Room, and Open Plan Office, suggesting the earlier participants reach a stable performance, the higher their FHR is. D. Performance for each carrier phrase length, d’ was significantly better as the ‘carrier phrases length’ increased except for CP2 vs. CP3 (n.s), where a roll over effect was observed. E. Performance to |Color| and |Number| for CP0 i.e., our proxy for short-term adaptation the carrier phrase where exposure to an environment remained minimal. Here all the environments have been collapsed i.e., rANOVA, main effect of “target word”. F-G. Hit Rate for Initial (1-2) and Steady (9-10) trials for keword |Color| in the CP0 condition, (F) in Reverberant conditions, indicating a significant improvement in performance for Steady trials, likely due to the accumulation of acoustic/environmental knowledge i.e., meta-adaptation (G) in anechoic (no echoes) showing a lack of improvement in performance i.e., lack of meta-adaptation in the absence of reverberation. https://doi.org/10.25949/24295342.v1

Learning effects in ‘sham’ and ‘TMS’ conditions.

A. Schematic representation of functional connections between the dorsolateral prefrontal cortex (dlPFC) and primary auditory cortex (A1) under transcranial magnetic stimulation (TMS). B. Average exposure time (s) ‘sham TMS’ and ‘real TMS’ conditions. Correct responses across time spent in each room for ‘sham’ and ‘TMS’ conditions. Solid curves in pink and gray correspond to mean data from ‘TMS’ and ‘Sham’ conditions respectively (after 5-time point moving averages, equivalent to ∼7s, were applied), with the associated shaded areas representing standard errors of the mean. Dashed black (sham TMS) and pink (real TMS) lines represent the time point at which participants reached a stable performance in each environment [i.e., ± 10% of the Final Hit Rate (FHR)], here ‘global learning’, no statistical differences were found for ‘global learning’ between TMS conditions. C. Shows a correlation analysis between FHR and the time where 10% of the FHR is achieved in under TMS (pink) and Sham (gray) conditions. Significant negative correlations (Pearson) were observed under Sham stimulation, suggesting the earlier participants reach a stable performance, the higher their FHR is. This relationship was disrupted (lack of significant correlation) under TMS conditions. D. Overall Performance in TMS and Sham exposed participants i.e., collapsed sensitivity (d’) to |Color| and |Number| for all carrier phrases and environments (significant main effect of TMS conditions: Univariate ANOVA). d’ values plotted in bright pink corresponds to the ‘real TMS’ condition whereas the ‘sham TMS’ accuracy is plotted in gray. Mean d’ is denoted as circles in the boxplot (n=11 for ‘sham TMS’ conditions and n=10 for ‘real TMS’ conditions). E. Significant Interaction ‘TMS condition’ x ‘carrier phrase length’. d’ was significantly better as the ‘carrier phrases length’ increased for ‘sham TMS’ compared to ‘real TMS’ except for CPO (n.s), i.e., our proxy for short-term adaptation and minimal exposure to acoustic environments. F. Performance to |Color| and |Number| for CP0 i.e., our proxy for short-term adaptation and the carrier phrase where exposure to an environment remained minimal. Here all the environments have been collapsed as only a main effect of “target word”, performance to number always significantly higher than for colour independent of TMS condition. G. Schematic for assessing short-term adaptation (performance to |Number| in the Initial Trials) and meta-adaptation (performance to |Color| between Initial and Steady Trials). H. Isolated short-term adaptation effects for performance to |Number| in CP0. Hit Rate for Initial (1-2) trials only for |Number| in the CP0 condition for both TMS (pink) and Sham (gray) conditions. Notice that no differences were observed between performance |Number| under Sham or TMS conditions, suggesting no disruption of short-term adaptation.I. Hit Rate for Initial (1-2) and Steady (9-10) trials only for |Color| in the CP0 condition for both TMS (pink) and Sham (gray) conditions. Notice that only a significant improvement in performance for from Intitial to Steady trials was observed in the ‘sham TMS’ condition. https://doi.org/10.25949/24295342.v1

Performance in different 3-room combinations where acoustics of 3 rooms span an ecological range.

A. Overall hit rate for different 3-room combinations. Mean data for four different 3-room combinations are plotted where Open-Plan Office (dark blue) and Underground Car Park (yellow) were always included and the context room was either Anechoic (black), Living Room (red), Lecture Room (green), Lecture Room (green) or Highly Reflective (purple). B. Performance in Open-Plan Office and Car Park when paired with different context rooms. Final hit rates were highest for Office and Car Park when presented in conjunction with the Lecture Room, followed by the Living Room and Anechoic with poorest performance observed when the Highly Reflective space was the context. C. Environments tested compared to ecological range. The RT60traers for the five test rooms, calculated as the median RT60 value across 31 frequency sub-bands (Traer & McDermott, 2016), are plotted (colored boxes) above a histogram of the median RT60traers calculated for 199 indoor (filled bars) and 72 outdoor (empty bars) spaces recorded by Traer and McDermott (2016). D. Frequency dependence of reverberation time (RT60traer) in test rooms compared to ecological range. RT60s of the 5 test rooms are displayed for frequency sub-bands used to calculate RT60traer (colored lines). Quartiles for combined indoor and outdoor spaces (Traer & McDermott, 2016) are plotted as black dashed lines. Decreasing reverberation time above 0.5kHz is observed for Living Room, Lecture Room and Car Park as has been typically described for indoor spaces (Traer & McDermott, 2016); however RT60traer profiles for Open-Plan Office and Highly Reflective space were notable for their longer reverberation times at and above 1kHz. https://doi.org/10.25949/24295342.v1

Pairwise comparisons between performance (d’) for carrier phrases length n 22 participants.

Pairwise comparisons among performance (d’) for Interaction ‘target word’ x ‘carrier phrase length’ in 22 participants

Pairwise comparisons among performance (Final Hit Rate) for the six talkers in 22 participants.

Pairwise comparisons between performance (Final Hit Rate) for carrier phrases length in 22 participants anechoic conditions.

Pairwise comparisons for interaction ‘TMS condition’ x ‘carrier phrase length’.

Time courses of performance for 22 subjects in lecture theatre (green), open plan office (blue) and car park (yellow).

Solid lines represent actual cumulative hit rate data for each room that were calculated from raw data using a 5-point moving average (equivalent to ∼7s). Open circles represents best fit of double exponential with adjusted R2 values shown above for each individual. https://doi.org/10.25949/24295342.v1

Comparison of Final Hit Rate (FHR) performances for different talkers.

A. Performance, calculated as FHR in all rooms, is shown for the 6 different talkers (1-3 male and 4-6 female) presented to 22 individuals who did not undergo experimental TMS during the task involving the Lecture Room/Open-Plan Office and Car Park RIRs. B. Performance was also similarly calculated for the different talkers but is now separated by the room RIR associated with the specific talker. https://doi.org/10.25949/24295342.v1

Hit Rate performances for |Color| and |Number| independent of environment presented or carrier phrase length.

A. Performance, calculated for the 1262 trials where 22 participants heard in 2 consecutive trials the same talker, a Wilcoxon signed rank test revealed no significant differences in performance between trial 1 vs. trial 2 (n=1262, Z= −0.42, p=0.68). B. Performance was also similarly calculated when listeners heard the same talker in three consecutive trials, this happened in a total of 217 trials and a Wilcoxon signed rank test revealed no significant differences in performance between trial 1 vs. trial 3 (n=217, Z= −0.17, p=0.87). C. Performance calculated in 40 trials were listeners heard the same talker in four consecutive trials. No statistical differences were observed between Trial 1 and Trial 4 (n=40, Z= −0.159, p=0.11).

Time courses of performance for 11 sham-TMS subjects in lecture theatre (green), open plan office (blue) and car park (yellow).

Solid lines represent actual cumulative hit rate data for each room that were calculated from raw data using a 5-point moving average (equivalent to ∼7s). Open circles represents best fit of double exponential with adjusted R2 values shown above for each individual. https://doi.org/10.25949/24295342.v1

Time courses of performance for 10 TMSin lecture theatre (green), open plan office (blue) and car park (yellow).

Solid lines represent actual cumulative hit rate data for each room that were calculated from raw data using a 5-point moving average (equivalent to ∼7s). Open circles represents best fit of double exponential with adjusted R2 values shown above for each individual. https://doi.org/10.25949/24295342.v1