diff --git a/CHANGELOG.md b/CHANGELOG.md
index 31d1a1ff5..408bb055e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1864,6 +1864,25 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
   Changed). Both warn on use and are removed in 5.0, one release after the
   3.x aliases.
 
+### Removed
+
+- The flat pre-3.2 module paths, as 3.2 announced when it deprecated them:
+  `phonometry.levels`, `phonometry.insulation`, `phonometry.room_ir` and the
+  eighty-odd siblings that the modularization grouped into domain subpackages
+  raise `ModuleNotFoundError` now, as does `phonometry.loudness`, deprecated
+  since 3.1. Only the module paths went. The names they held are still
+  exported flat, so `from phonometry import leq` reads as it always has, and
+  so does every other name in `phonometry.__all__`. The 4.0 taxonomy aliases
+  are untouched and stay until 5.0.
+
+  A pickle written by a 3.x release carries a flat path in the `__module__` of
+  its classes, and those no longer resolve; one written by 3.2 or later
+  through a subpackage path still does, until 5.0.
+
+  With one generation left, `_compat.py` keeps one table, and the alias
+  machinery no longer carries the per-entry wording overrides that the 3.1
+  deprecations needed.
+
 ### Fixed
 
 - The committed example `.report()` fiches were not checked against the code
diff --git a/docs/api-reference.md b/docs/api-reference.md
index 32559be1c..2a9149a1d 100644
--- a/docs/api-reference.md
+++ b/docs/api-reference.md
@@ -44,15 +44,14 @@ contour = aircraft.noise_contour(path, powers, distances, sel, lmax, x=gx, y=gy)
 | `phonometry.simulation` | 2D acoustic FDTD wave simulation (staggered grid, sources, probes, impedance boundaries, obstacles) |
 
 Every name in the table below is also re-exported at the top level, so
-`from phonometry import <name>` works for every row. Two generations of module
-paths are still importable and warn on use. The pre-3.2 flat ones, written
-without a subpackage, are removed in 4.0; the pre-4.0 ones that the taxonomy
-moved are removed in 5.0. Reading a moved name from the namespace it left
-warns and delegates as well.
+`from phonometry import <name>` works for every row. The pre-4.0 module paths
+that the taxonomy moved are still importable and warn on use; they are removed
+in 5.0. Reading a moved name from the namespace it left warns and delegates as
+well. The pre-3.2 flat paths, written without a subpackage, were removed in
+4.0 as announced.
 
 | Deprecated path | Resolves to | Removed in |
 | :--- | :--- | :--- |
-| `phonometry.insulation` | `phonometry.building.measurement.insulation` | 4.0 |
 | `phonometry.metrology.levels` | `phonometry.signals.levels` | 5.0 |
 | `phonometry.hearing.sti` | `phonometry.speech.sti` | 5.0 |
 | `phonometry.vibration.human_vibration` | `phonometry.vibration.human.exposure` | 5.0 |
diff --git a/llms-full.txt b/llms-full.txt
index eecb24982..d7b08f935 100644
--- a/llms-full.txt
+++ b/llms-full.txt
@@ -2257,15 +2257,14 @@ contour = aircraft.noise_contour(path, powers, distances, sel, lmax, x=gx, y=gy)
 | `phonometry.simulation` | 2D acoustic FDTD wave simulation (staggered grid, sources, probes, impedance boundaries, obstacles) |
 
 Every name in the table below is also re-exported at the top level, so
-`from phonometry import <name>` works for every row. Two generations of module
-paths are still importable and warn on use. The pre-3.2 flat ones, written
-without a subpackage, are removed in 4.0; the pre-4.0 ones that the taxonomy
-moved are removed in 5.0. Reading a moved name from the namespace it left
-warns and delegates as well.
+`from phonometry import <name>` works for every row. The pre-4.0 module paths
+that the taxonomy moved are still importable and warn on use; they are removed
+in 5.0. Reading a moved name from the namespace it left warns and delegates as
+well. The pre-3.2 flat paths, written without a subpackage, were removed in
+4.0 as announced.
 
 | Deprecated path | Resolves to | Removed in |
 | :--- | :--- | :--- |
-| `phonometry.insulation` | `phonometry.building.measurement.insulation` | 4.0 |
 | `phonometry.metrology.levels` | `phonometry.signals.levels` | 5.0 |
 | `phonometry.hearing.sti` | `phonometry.speech.sti` | 5.0 |
 | `phonometry.vibration.human_vibration` | `phonometry.vibration.human.exposure` | 5.0 |
diff --git a/scripts/generate_api_docs.py b/scripts/generate_api_docs.py
index 145999e76..b507ef76d 100644
--- a/scripts/generate_api_docs.py
+++ b/scripts/generate_api_docs.py
@@ -774,38 +774,9 @@ def build_model() -> tuple[list[ModuleDoc], dict[str, str], list[str]]:
     if len(slugs) != len(pages):
         raise ValueError("slug collision between module pages")
 
-    # Docstrings still reference the flat pre-modularization module names
-    # (":mod:`phonometry.insulation`"); alias them to the current pages.
-    labels: dict[str, list[str]] = {}
-    for page in pages:
-        if page.module != "phonometry":
-            labels.setdefault(page.label, []).append(page.url)
-    for label, urls in labels.items():
-        alias = f"phonometry.{label}"
-        if alias not in xref and len(urls) == 1:
-            xref[alias] = urls[0]
-    for old, new in _LEGACY_MODULE_ALIASES.items():
-        if old not in xref and new in xref:
-            xref[old] = xref[new]
     return pages, xref, issues
 
 
-#: Flat pre-modularization module names whose basename also changed; the
-#: unchanged ones are aliased automatically from the page basenames above.
-_LEGACY_MODULE_ALIASES = {
-    "phonometry.aircraft_atmospheric_absorption": (
-        "phonometry.aircraft.atmospheric_absorption"
-    ),
-    "phonometry.environmental_measurement": (
-        "phonometry.environment.assessment.measurement"
-    ),
-    "phonometry.underwater_acoustics": "phonometry.underwater.acoustics",
-    "phonometry.underwater_propagation":
-        "phonometry.underwater.propagation.closed_form",
-    "phonometry.underwater_sound_speed": "phonometry.underwater.propagation.sound_speed",
-}
-
-
 _TOP_LEVEL_INTRO = (
     "Package-level names defined in `phonometry/__init__.py` itself. "
     "Every public name in the library can also be imported directly from "
diff --git a/site/src/content/docs/reference/api/aeroacoustics/atmospheric-absorption.md b/site/src/content/docs/reference/api/aeroacoustics/atmospheric-absorption.md
index 6b818185b..2ddff3c3f 100644
--- a/site/src/content/docs/reference/api/aeroacoustics/atmospheric-absorption.md
+++ b/site/src/content/docs/reference/api/aeroacoustics/atmospheric-absorption.md
@@ -11,7 +11,7 @@ Aircraft noise certification (14 CFR Part 36, ICAO Annex 16 Vol. I) works with
 one-third-octave-band spectra, and correcting a measured flyover to reference
 atmospheric conditions requires the band attenuation over the propagation path.
 The pure-tone attenuation coefficient is the ISO 9613-1 one (identical, per ARP
-5534 §3.1) already provided by [`phonometry.air_absorption.air_attenuation`](/phonometry/reference/api/environment/air-absorption/#air_attenuation);
+5534 §3.1) already provided by [`air_attenuation`](/phonometry/reference/api/environment/air-absorption/#air_attenuation);
 this module adds the **SAE Method** (ARP 5534 §3.2.2), a regression that turns
 the pure-tone mid-band path-length attenuation into the one-third-octave-band
 attenuation and stays consistent with the ISO/ANSI Exact Method well beyond the
diff --git a/site/src/content/docs/reference/api/electroacoustics/swept-sine.md b/site/src/content/docs/reference/api/electroacoustics/swept-sine.md
index 59a0d3c10..079ce9c04 100644
--- a/site/src/content/docs/reference/api/electroacoustics/swept-sine.md
+++ b/site/src/content/docs/reference/api/electroacoustics/swept-sine.md
@@ -42,7 +42,7 @@ Two excitation/deconvolution pairs are implemented:
 * **Exponential sweep with the Farina inverse filter** -- the classical
   ESS of [`phonometry.sweep_signal`](/phonometry/reference/api/rooms/impulse-response/#sweep_signal) deconvolved by the time-reversed,
   amplitude-compensated sweep (Farina 2000; ISO 18233:2006 Figure B.2,
-  the machinery of [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/)). The harmonic
+  the machinery of [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/#impulse_response)). The harmonic
   *magnitudes* are correct, but the sweep is not synchronized: its `-1`
   phase term breaks the time-shift/harmonic equivalence, so the phases of
   `H_n` for $n \ge 2$ depend on the excitation and are not
diff --git a/site/src/content/docs/reference/api/environment/air-absorption.md b/site/src/content/docs/reference/api/environment/air-absorption.md
index 37d7ca75c..4631de373 100644
--- a/site/src/content/docs/reference/api/environment/air-absorption.md
+++ b/site/src/content/docs/reference/api/environment/air-absorption.md
@@ -55,7 +55,7 @@ $f_m = 1000 \cdot 10^{k/10}$, `k` integer. Pass
 `exact_midband=True` to snap the requested frequencies onto that grid and
 reproduce Table 1 exactly.
 
-This module closes the loop with [`phonometry.materials.absorbers.sound_absorption`](/phonometry/reference/api/materials/sound-absorption/) (ISO 354),
+This module closes the loop with [`sound_absorption`](/phonometry/reference/api/materials/sound-absorption/) (ISO 354),
 whose air power-attenuation coefficient `m` (1/m) is defined only through
 the ISO 9613-1 `alpha` via $m = \alpha / (10 \log_{10} e)$.
 [`air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m) returns that `m` directly.
@@ -99,7 +99,7 @@ scalars.
 ISO 354:2003 defers its air power-attenuation coefficient `m` (1/m)
 entirely to this `alpha` via $m = \alpha / (10 \log_{10} e)$. Use
 [`air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m) to obtain that `m` for
-[`phonometry.sound_absorption.absorption_area`](/phonometry/reference/api/materials/sound-absorption/#absorption_area) /
+[`absorption_area`](/phonometry/reference/api/materials/sound-absorption/#absorption_area) /
 [`absorption_coefficient`](/phonometry/reference/api/materials/sound-absorption/#absorption_coefficient).
 :::
 
@@ -121,7 +121,7 @@ ISO 354 air power-attenuation coefficient `m` (1/m) from conditions.
 Convenience composition of [`air_attenuation`](/phonometry/reference/api/environment/air-absorption/#air_attenuation) (ISO 9613-1 `alpha` in
 dB/m) with the ISO 354:2003 (8.1.2.1) conversion
 $m = \alpha / (10 \log_{10} e)$
-(via [`phonometry.sound_absorption.attenuation_from_alpha`](/phonometry/reference/api/materials/sound-absorption/#attenuation_from_alpha)). It lets an
+(via [`attenuation_from_alpha`](/phonometry/reference/api/materials/sound-absorption/#attenuation_from_alpha)). It lets an
 ISO 354 caller feed real atmospheric conditions into
 [`absorption_area`](/phonometry/reference/api/materials/sound-absorption/#absorption_area) /
 [`absorption_coefficient`](/phonometry/reference/api/materials/sound-absorption/#absorption_coefficient) instead of
diff --git a/site/src/content/docs/reference/api/environment/outdoor-propagation.md b/site/src/content/docs/reference/api/environment/outdoor-propagation.md
index 5373b07c9..b6d5f8491 100644
--- a/site/src/content/docs/reference/api/environment/outdoor-propagation.md
+++ b/site/src/content/docs/reference/api/environment/outdoor-propagation.md
@@ -30,7 +30,7 @@ Implemented here are the four general terms of clause 7:
 
 * `Adiv` geometrical divergence, $20 \log_{10}(d/d_0) + 11$ (Eq. (7));
 * `Aatm` atmospheric absorption, $\alpha d$ (Eq. (8)) with `alpha`
-  the ISO 9613-1 coefficient supplied by [`phonometry.environment.propagation.air_absorption`](/phonometry/reference/api/environment/air-absorption/);
+  the ISO 9613-1 coefficient supplied by [`air_absorption`](/phonometry/reference/api/environment/air-absorption/);
 * `Agr` ground effect, both the general per-region method of 7.3.1 with the
   Table 3 functions `a'/b'/c'/d'` (Eq. (9)) and the alternative simplified
   method of 7.3.2 (Eq. (10));
@@ -66,7 +66,7 @@ Attenuation due to atmospheric absorption (ISO 9613-2:1996, Eq. (8)).
 
 $A_{atm} = \alpha d$ with `alpha` the ISO 9613-1 atmospheric
 attenuation
-coefficient (here in dB/m, from [`phonometry.air_absorption.air_attenuation`](/phonometry/reference/api/environment/air-absorption/#air_attenuation))
+coefficient (here in dB/m, from [`air_attenuation`](/phonometry/reference/api/environment/air-absorption/#air_attenuation))
 at each octave-band midband frequency. Eq. (8) writes `alpha` in dB/km
 with $A_{atm} = \alpha_{\text{dB/km}} \, d / 1000$; the two forms
 are identical.
diff --git a/site/src/content/docs/reference/api/environment/wind-turbine.md b/site/src/content/docs/reference/api/environment/wind-turbine.md
index 094f76ffc..fc4407972 100644
--- a/site/src/content/docs/reference/api/environment/wind-turbine.md
+++ b/site/src/content/docs/reference/api/environment/wind-turbine.md
@@ -20,7 +20,7 @@ Two closed-form quantities of the standard:
   `ΔL_a` that decides whether a tone is audible.
 
 The tonal-audibility formula itself is the ISO 1996-2 Annex C one already in
-[`phonometry.environment.assessment.measurement`](/phonometry/reference/api/environment/measurement/); what is specific to IEC 61400-11 is
+[`measurement`](/phonometry/reference/api/environment/measurement/); what is specific to IEC 61400-11 is
 how the tone and masking-noise levels and the (Zwicker) critical band are
 determined from the narrowband spectrum. The rating adjustment `K_T` is the
 ISO 1996-2 [`tonal_adjustment`](/phonometry/reference/api/environment/measurement/#tonal_adjustment). The
@@ -49,7 +49,7 @@ pressure doubling; $S_0 = 1$ m².
 
 | Name | Description |
 | :--- | :--- |
-| `band_levels` | Background-corrected A-weighted band sound pressure levels `L_p,i`, in dB (scalar or per band). The 61400-11-specific background correction (Formula 23 subtraction with the 3-6 dB asterisk marking and the \<= 3 dB not-reported rule, subclause 9.3) is out of scope here and must be applied beforehand; note its rule set differs from the ISO 1996-2 correction in [`phonometry.environment.assessment.measurement.residual_sound_correction`](/phonometry/reference/api/environment/measurement/#residual_sound_correction). |
+| `band_levels` | Background-corrected A-weighted band sound pressure levels `L_p,i`, in dB (scalar or per band). The 61400-11-specific background correction (Formula 23 subtraction with the 3-6 dB asterisk marking and the \<= 3 dB not-reported rule, subclause 9.3) is out of scope here and must be applied beforehand; note its rule set differs from the ISO 1996-2 correction in [`residual_sound_correction`](/phonometry/reference/api/environment/measurement/#residual_sound_correction). |
 | `r1` | Slant distance `R1` to the rotor centre, in m. |
 
 **Returns:** The apparent sound power level `L_WA`, in dB re 1 pW.
diff --git a/site/src/content/docs/reference/api/materials/road-absorption.md b/site/src/content/docs/reference/api/materials/road-absorption.md
index 25fc3f7f9..adc4f76d6 100644
--- a/site/src/content/docs/reference/api/materials/road-absorption.md
+++ b/site/src/content/docs/reference/api/materials/road-absorption.md
@@ -37,7 +37,7 @@ opposite ends of the absorption scale and are **not** interchangeable:
   (measured `alpha` below ~0.15). Part 2 does not restate the
   transfer-function / reflection-factor / absorption mathematics; Clauses 4,
   5.7 and 6.6 defer it to ISO 10534-2. The core computation therefore lives in
-  [`phonometry.impedance_tube.two_microphone_impedance`](/phonometry/reference/api/materials/impedance-tube/#two_microphone_impedance) and is **not**
+  [`two_microphone_impedance`](/phonometry/reference/api/materials/impedance-tube/#two_microphone_impedance) and is **not**
   reimplemented here. This module contributes only the Part-2 tube
   geometry/validity helpers (upper usable frequency, microphone-spacing bounds,
   the 250-1600 Hz one-third-octave working range) and the Annex A internal-loss
diff --git a/site/src/content/docs/reference/api/rooms/acoustics.md b/site/src/content/docs/reference/api/rooms/acoustics.md
index ff60ab78c..1a045f9dd 100644
--- a/site/src/content/docs/reference/api/rooms/acoustics.md
+++ b/site/src/content/docs/reference/api/rooms/acoustics.md
@@ -69,7 +69,7 @@ integrated impulse response, Clause 6).
 
 | Name | Description |
 | :--- | :--- |
-| `ir` | Measured impulse response (1D), e.g. from [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/) (ISO 18233). |
+| `ir` | Measured impulse response (1D), e.g. from [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/#impulse_response) (ISO 18233). |
 | `fs` | Sample rate in Hz. |
 | `band` | Optional band centre frequency in Hz. When given, the impulse response is first filtered with the matching IEC 61260 fractional-octave filter; when None the broadband response is integrated directly. |
 | `fraction` | Bandwidth fraction of the band filter (1 = octave, 3 = one-third octave). Only used when `band` is not None. |
diff --git a/site/src/content/docs/reference/api/rooms/image-source.md b/site/src/content/docs/reference/api/rooms/image-source.md
index 51caa02b2..546fb0971 100644
--- a/site/src/content/docs/reference/api/rooms/image-source.md
+++ b/site/src/content/docs/reference/api/rooms/image-source.md
@@ -144,7 +144,7 @@ reproduces the Eyring reverberation time of the room (Kuttruff Equation
 | `fs` | Sample rate, Hz. |
 | `max_order` | Reflection-order cut-off (total wall reflections). The shoebox has $(2/3)(2 i_0^3 + 3 i_0^2 + 4 i_0)$ audible images up to order `i0` (Kuttruff Equation (9.23)). Default 20. |
 | `speed_of_sound` | Speed of sound `c`, m/s (default [`DEFAULT_SPEED_OF_SOUND`](/phonometry/reference/api/materials/road-absorption/#default_speed_of_sound)). |
-| `air_attenuation` | Air *intensity* attenuation constant `m`, in neper per metre (scalar or per-band); the pressure amplitude of each path is scaled by $e^{-m r / 2}$ (Kuttruff 4.1). Default 0 (air absorption neglected). Obtain a physical `m` from [`phonometry.air_absorption.air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m). |
+| `air_attenuation` | Air *intensity* attenuation constant `m`, in neper per metre (scalar or per-band); the pressure amplitude of each path is scaled by $e^{-m r / 2}$ (Kuttruff 4.1). Default 0 (air absorption neglected). Obtain a physical `m` from [`air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m). |
 | `duration` | RIR length, s; default the latest image arrival rounded up to the next sample. |
 | `frequencies` | Optional band centre frequencies, Hz, labelling a per-band result. When given, its length must match the band count of `absorption` (or broadcast against it). |
 
diff --git a/site/src/content/docs/reference/api/rooms/reverberation-prediction.md b/site/src/content/docs/reference/api/rooms/reverberation-prediction.md
index 37ba98ba2..180b7271c 100644
--- a/site/src/content/docs/reference/api/rooms/reverberation-prediction.md
+++ b/site/src/content/docs/reference/api/rooms/reverberation-prediction.md
@@ -47,7 +47,7 @@ default $c_0 = 343$ m/s it takes the familiar textbook value
 Air absorption enters every model through the `air_attenuation` power
 coefficient `m` (in neper per metre) as the additive term $4 m V$;
 obtain a physical `m` from temperature and humidity with
-[`phonometry.air_absorption.air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m).
+[`air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m).
 
 Each model enforces its own mathematical domain on the absorption
 coefficients. Sabine's linear sum is finite for any non-negative coefficient,
@@ -390,7 +390,7 @@ error (a percentage passed instead of a fraction).
 | :--- | :--- |
 | `volume` | Room volume `V`, m3. |
 | `surfaces` | Sequence of `(area, absorption_coefficient)` pairs; each coefficient a scalar or a per-band array in `[0, 2]`. |
-| `air_attenuation` | Air power-attenuation coefficient `m`, in neper per metre (scalar or per-band); see [`phonometry.air_absorption.air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m). Default `0` (air absorption neglected). |
+| `air_attenuation` | Air power-attenuation coefficient `m`, in neper per metre (scalar or per-band); see [`air_attenuation_m`](/phonometry/reference/api/environment/air-absorption/#air_attenuation_m). Default `0` (air absorption neglected). |
 | `speed_of_sound` | Speed of sound `c0`, m/s (default [`DEFAULT_SPEED_OF_SOUND`](/phonometry/reference/api/materials/road-absorption/#default_speed_of_sound), giving the factor `0.161`). |
 
 **Returns:** The reverberation time `T`, s; a float for scalar inputs, otherwise a per-band array.
diff --git a/site/src/content/docs/reference/api/signals/inversion.md b/site/src/content/docs/reference/api/signals/inversion.md
index ff9ae8d95..a9bbf1010 100644
--- a/site/src/content/docs/reference/api/signals/inversion.md
+++ b/site/src/content/docs/reference/api/signals/inversion.md
@@ -176,7 +176,7 @@ Sec. 2.4).
 
 Both regularization levels are *relative* to the peak of
 $\lvert H \rvert^2$
-(like the scalar `regularization` of [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/),
+(like the scalar `regularization` of [`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/#impulse_response),
 which this generalises): in-band the equalized magnitude deviates from
 unity by at most `regularization_inside * max|H|**2 / min|H|**2` --
 the analytic residue
diff --git a/site/src/content/docs/reference/api/signals/phase.md b/site/src/content/docs/reference/api/signals/phase.md
index 7827eb799..ac654c8c8 100644
--- a/site/src/content/docs/reference/api/signals/phase.md
+++ b/site/src/content/docs/reference/api/signals/phase.md
@@ -30,7 +30,7 @@ phase.
 Sampling precautions (documented, and the reason for the `oversample`
 padding): the estimate operates on a *uniformly sampled* one-sided response
 (DC to Nyquist inclusive, the layout of `numpy.fft.rfft` and of
-[`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/) spectra). The real cepstrum of the
+[`phonometry.impulse_response`](/phonometry/reference/api/rooms/impulse-response/#impulse_response) spectra). The real cepstrum of the
 sampled log-magnitude is time-aliased when the grid is coarse relative to
 how sharp the response is, so the magnitude is resampled onto an
 `oversample` times denser grid by exact trigonometric (zero-padded
diff --git a/src/phonometry/__init__.py b/src/phonometry/__init__.py
index 8847e6484..d1c97e577 100644
--- a/src/phonometry/__init__.py
+++ b/src/phonometry/__init__.py
@@ -2851,7 +2851,7 @@ def __getattr__(name: str) -> Any:
     return globals()[canonical]
 
 
-# Deprecated module-path aliases for the 3.2 package reorganization: importing
-# the package installs sys.modules shims for every moved public module (see
-# phonometry/_compat.py; removed in 4.0).
+# Deprecated module-path aliases for the 4.0 taxonomy: importing the package
+# installs sys.modules shims for every moved public module (see
+# phonometry/_compat.py; removed in 5.0).
 from . import _compat as _compat
diff --git a/src/phonometry/_compat.py b/src/phonometry/_compat.py
index 2678f8bfd..10bc9beef 100644
--- a/src/phonometry/_compat.py
+++ b/src/phonometry/_compat.py
@@ -1,20 +1,20 @@
 #  Copyright (c) 2026. Jose Manuel Requena Plens
 """Deprecated module-path aliases for the phonometry package layout.
 
-Two generations of aliases live here, each with its own removal date:
-
-* :data:`_MOVED_3X` covers the 3.2 modularization, which grouped the flat
-  top-level modules into domain subpackages (``phonometry.building``,
-  ``phonometry.underwater``, ...). Removed in 4.0. Its targets follow the
-  modules wherever they land, so a 3.x path always resolves in one hop.
-* :data:`_MOVED_4X` covers the 4.0 taxonomy, which splits the oversized
-  subpackages into domain ones (``phonometry.metrology`` into
-  ``phonometry.filters``, ``phonometry.signals`` and a narrowed
-  ``phonometry.metrology``; the speech intelligibility of
-  ``phonometry.hearing`` into ``phonometry.speech``), gives the large domains
-  a second level (``phonometry.vibration`` into ``structural``, ``human`` and
-  ``machinery``) and renames ``phonometry.environmental`` to
-  ``phonometry.environment``. Removed in 5.0.
+:data:`_MOVED_4X` covers the 4.0 taxonomy, which splits the oversized
+subpackages into domain ones (``phonometry.metrology`` into
+``phonometry.filters``, ``phonometry.signals`` and a narrowed
+``phonometry.metrology``; the speech intelligibility of
+``phonometry.hearing`` into ``phonometry.speech``), gives the large domains a
+second level (``phonometry.vibration`` into ``structural``, ``human`` and
+``machinery``) and renames ``phonometry.environmental`` to
+``phonometry.environment``. Removed in 5.0.
+
+The 3.2 generation, which grouped the flat top-level modules into domain
+subpackages, announced 4.0 as its removal and is gone: ``phonometry.levels``
+and its eighty-odd siblings raise ``ModuleNotFoundError`` now. Only the module
+paths went: the names they held are still exported flat, so ``from phonometry
+import leq`` reads as it always has.
 
 Every public module path that moved stays importable through the shims
 registered here: ``import phonometry.<old>`` and ``from phonometry.<old>
@@ -27,9 +27,6 @@
 documentation leads with. :func:`_namespace_shim` keeps those attribute reads
 working from the namespace they left, with the same notice.
 
-This generalizes the former ``phonometry.loudness`` PEP 562 shim (that module
-file is gone; its entry lives in the table below with its original 3.1 wording
-preserved).
 """
 
 from __future__ import annotations
@@ -42,134 +39,10 @@
 
 from ._internal.warnings import _warn_renamed
 
-#: The one module NT ACOU 112 and ISO/PAS 1996-3 share since 4.0. Four old
-#: paths land on it, from both generations of the table.
+#: The one module NT ACOU 112 and ISO/PAS 1996-3 share since 4.0. Three old
+#: paths land on it.
 _IMPULSIVE_SOUND = "phonometry.environment.assessment.impulsive_sound"
 
-#: Old public module path -> relocated module path. One row per moved module.
-_MOVED_3X: dict[str, str] = {
-    "phonometry.utils": "phonometry._internal.utils",
-    "phonometry._warnings": "phonometry._internal.warnings",
-    "phonometry.calibration": "phonometry.metrology.calibration",
-    "phonometry.compliance": "phonometry.filters.compliance",
-    "phonometry.core": "phonometry.filters.core",
-    "phonometry.filter_design": "phonometry.filters.design",
-    "phonometry.frequencies": "phonometry.filters.frequencies",
-    "phonometry.levels": "phonometry.signals.levels",
-    "phonometry.parametric_filters": "phonometry.filters.weighting",
-    "phonometry.uncertainty": "phonometry.metrology.uncertainty",
-    "phonometry.fluctuation_strength": "phonometry.psychoacoustics.quality.fluctuation_strength",
-    "phonometry.loudness_contours": "phonometry.psychoacoustics.loudness.contours",
-    "phonometry.loudness_ecma": "phonometry.psychoacoustics.loudness.ecma",
-    "phonometry.loudness_moore_glasberg": "phonometry.psychoacoustics.loudness.moore_glasberg",
-    "phonometry.loudness_moore_glasberg_time": "phonometry.psychoacoustics.loudness.moore_glasberg_time",
-    "phonometry.loudness_zwicker": "phonometry.psychoacoustics.loudness.zwicker",
-    "phonometry.psychoacoustic_annoyance": "phonometry.psychoacoustics.quality.annoyance",
-    "phonometry.roughness_ecma": "phonometry.psychoacoustics.quality.roughness_ecma",
-    "phonometry.sharpness": "phonometry.psychoacoustics.quality.sharpness",
-    "phonometry.tonality": "phonometry.psychoacoustics.quality.tonality",
-    "phonometry.tonality_ecma": "phonometry.psychoacoustics.quality.tonality_ecma",
-    "phonometry.tone_audibility": "phonometry.psychoacoustics.quality.tone_audibility",
-    "phonometry.noise_induced_hearing_loss": "phonometry.hearing.noise_induced_hearing_loss",
-    "phonometry.occupational_exposure": "phonometry.hearing.occupational_exposure",
-    "phonometry.sii": "phonometry.speech.sii",
-    "phonometry.sti": "phonometry.speech.sti",
-    "phonometry.intensity": "phonometry.emission.intensity",
-    "phonometry.sound_power": "phonometry.emission.sound_power",
-    "phonometry.sound_power_intensity": "phonometry.emission.sound_power_intensity",
-    "phonometry.sound_power_reverberation": "phonometry.emission.sound_power_reverberation",
-    "phonometry.vibration_sound_power": "phonometry.emission.vibration_sound_power",
-    "phonometry.absorption_rating":
-        "phonometry.materials.absorbers.rating",
-    "phonometry.absorption_uncertainty":
-        "phonometry.materials.absorbers.uncertainty",
-    "phonometry.airflow_resistance":
-        "phonometry.materials.absorbers.airflow_resistance",
-    "phonometry.dynamic_stiffness":
-        "phonometry.materials.resilient.dynamic_stiffness",
-    "phonometry.impedance_tube":
-        "phonometry.materials.absorbers.impedance_tube",
-    "phonometry.road_absorption":
-        "phonometry.materials.surfaces.road_absorption",
-    "phonometry.scattering_diffusion":
-        "phonometry.materials.diffusers.scattering_diffusion",
-    "phonometry.sound_absorption":
-        "phonometry.materials.absorbers.sound_absorption",
-    "phonometry.enclosed_space_absorption": "phonometry.room.enclosed_space_absorption",
-    "phonometry.open_plan": "phonometry.room.open_plan",
-    "phonometry.reverberation_prediction": "phonometry.room.reverberation_prediction",
-    "phonometry.room_acoustics": "phonometry.room.acoustics",
-    "phonometry.room_ir": "phonometry.room.impulse_response",
-    "phonometry.room_noise": "phonometry.room.noise_criteria",
-    "phonometry.building_prediction":
-        "phonometry.building.prediction.simplified_model",
-    "phonometry.building_uncertainty":
-        "phonometry.building.measurement.uncertainty",
-    "phonometry.facade_prediction":
-        "phonometry.building.prediction.facade",
-    "phonometry.flanking_transmission":
-        "phonometry.building.measurement.flanking_transmission",
-    "phonometry.floor_covering_improvement":
-        "phonometry.building.measurement.floor_covering_improvement",
-    "phonometry.installed_structure_borne":
-        "phonometry.building.prediction.installed_structure_borne",
-    "phonometry.insulation":
-        "phonometry.building.measurement.insulation",
-    "phonometry.intensity_insulation":
-        "phonometry.building.measurement.intensity_insulation",
-    "phonometry.lab_insulation":
-        "phonometry.building.measurement.lab_insulation",
-    "phonometry.structure_borne_power":
-        "phonometry.building.measurement.structure_borne_power",
-    "phonometry.survey_insulation":
-        "phonometry.building.measurement.survey_insulation",
-    "phonometry.human_vibration": "phonometry.vibration.human.exposure",
-    "phonometry.mechanical_mobility":
-        "phonometry.vibration.structural.mechanical_mobility",
-    "phonometry.multiple_shock_vibration":
-        "phonometry.vibration.human.multiple_shock",
-    "phonometry.transfer_stiffness":
-        "phonometry.vibration.structural.transfer_stiffness",
-    "phonometry.air_absorption":
-        "phonometry.environment.propagation.air_absorption",
-    "phonometry.environmental_measurement":
-        "phonometry.environment.assessment.measurement",
-    "phonometry.impulse_prominence": _IMPULSIVE_SOUND,
-    "phonometry.outdoor_propagation":
-        "phonometry.environment.propagation.outdoor_propagation",
-    "phonometry.wind_turbine_noise":
-        "phonometry.environment.sources.wind_turbine",
-    "phonometry.aircraft_atmospheric_absorption": "phonometry.aircraft.atmospheric_absorption",
-    "phonometry.aircraft_noise": "phonometry.aircraft.certification",
-    "phonometry.airport_noise": "phonometry.aircraft.airport_noise",
-    "phonometry.rotorcraft_noise": "phonometry.aircraft.rotorcraft_noise",
-    "phonometry.numerical_propagation": "phonometry.underwater.propagation.numerical",
-    "phonometry.ocean_ambient_noise": "phonometry.underwater.sources.ambient_noise",
-    "phonometry.pile_driving_noise": "phonometry.underwater.sources.pile_driving_noise",
-    "phonometry.seabed_reflection": "phonometry.underwater.propagation.seabed_reflection",
-    "phonometry.ship_radiated_noise": "phonometry.underwater.sources.ship_radiated_noise",
-    "phonometry.ship_traffic_noise": "phonometry.underwater.sources.ship_traffic_noise",
-    "phonometry.sonar_equation": "phonometry.underwater.sonar_equation",
-    "phonometry.underwater_acoustics": "phonometry.underwater.acoustics",
-    "phonometry.underwater_propagation":
-        "phonometry.underwater.propagation.closed_form",
-    "phonometry.underwater_sound_speed": "phonometry.underwater.propagation.sound_speed",
-    "phonometry.distortion": "phonometry.electroacoustics.distortion",
-    "phonometry.frequency_response": "phonometry.electroacoustics.frequency_response",
-    # <migrate:auto>
-}
-
-#: Entries whose deprecation predates 3.2 keep their original wording.
-_SINCE: dict[str, str] = {
-    "phonometry.loudness": "3.1",
-}
-
-#: Renames that were already shimmed before 3.2 (target differs from a plain
-#: package move). ``phonometry.loudness`` predates the reorganization and
-#: lands wherever the Zwicker module currently lives, so it reads its target
-#: from the row above instead of repeating it.
-_MOVED_3X["phonometry.loudness"] = _MOVED_3X["phonometry.loudness_zwicker"]
-
 #: Old module path -> relocated module path for the 4.0 taxonomy. The
 #: oversized ``metrology`` catch-all became three packages: the normalized
 #: frequency selectivity in ``filters``, the general signal analysis in
@@ -364,10 +237,9 @@
     # shadowing it is exactly what must not happen.
 }
 
-#: The two generations, each with the release that deprecated it and the one
-#: that removes it. Order matters only for readability; the paths are disjoint.
+#: The generations in force, each with the release that deprecated it and the
+#: one that removes it. The 3.2 generation was removed in 4.0, as announced.
 _GENERATIONS: tuple[tuple[dict[str, str], str, str], ...] = (
-    (_MOVED_3X, "3.2", "4.0"),
     (_MOVED_4X, "4.0", "5.0"),
 )
 
@@ -404,7 +276,7 @@ def __getattr__(name: str) -> Any:
         _warn_renamed(
             f"the '{old}' module",
             f"'{new}'",
-            since=_SINCE.get(old, since),
+            since=since,
             removed_in=removed_in,
         )
         return attr
@@ -541,9 +413,9 @@ def _install() -> None:
                 continue
             shim = _make_shim(old, new, since, removed_in)
             sys.modules[old] = shim
-            # `import phonometry.utils` also binds the attribute on the
-            # package; mirror that so `phonometry.utils` resolves without the
-            # import. Aliases below a subpackage are served by that package's
+            # `import phonometry.environmental` also binds the attribute on
+            # the package; mirror that so `phonometry.environmental` resolves
+            # without the import. Aliases below a subpackage are served by that package's
             # own shim (:func:`_namespace_shim`), which resolves the moved
             # public names first, so binding them here would shadow a function
             # with a module.
diff --git a/src/phonometry/_plot/psychoacoustics.py b/src/phonometry/_plot/psychoacoustics.py
index 54602ce4b..3f5a9f5c4 100644
--- a/src/phonometry/_plot/psychoacoustics.py
+++ b/src/phonometry/_plot/psychoacoustics.py
@@ -519,8 +519,8 @@ def plot_ecma_fluctuation_strength(
     ``ax`` is supplied only the time-dependent fluctuation strength is drawn
     on it and that single axes is returned.
 
-    :param result: An :class:`~phonometry.fluctuation_strength_ecma.
-        EcmaFluctuationStrength`.
+    :param result: An :class:`~phonometry.psychoacoustics.quality.
+        fluctuation_strength_ecma.EcmaFluctuationStrength`.
     :param ax: Existing axes to draw on, or ``None`` to create a figure.
     :param language: Label language, ``"en"`` (default) or ``"es"``.
     :param kwargs: Forwarded to the time-trace line ``plot`` call.
diff --git a/src/phonometry/aircraft/atmospheric_absorption.py b/src/phonometry/aircraft/atmospheric_absorption.py
index 5037bb0d9..b11bebd48 100644
--- a/src/phonometry/aircraft/atmospheric_absorption.py
+++ b/src/phonometry/aircraft/atmospheric_absorption.py
@@ -6,7 +6,7 @@
 one-third-octave-band spectra, and correcting a measured flyover to reference
 atmospheric conditions requires the band attenuation over the propagation path.
 The pure-tone attenuation coefficient is the ISO 9613-1 one (identical, per ARP
-5534 §3.1) already provided by :func:`phonometry.air_absorption.air_attenuation`;
+5534 §3.1) already provided by :func:`~phonometry.environment.propagation.air_absorption.air_attenuation`;
 this module adds the **SAE Method** (ARP 5534 §3.2.2), a regression that turns
 the pure-tone mid-band path-length attenuation into the one-third-octave-band
 attenuation and stays consistent with the ISO/ANSI Exact Method well beyond the
diff --git a/src/phonometry/building/measurement/uncertainty.py b/src/phonometry/building/measurement/uncertainty.py
index 9ac5b6f42..c0f5f2021 100644
--- a/src/phonometry/building/measurement/uncertainty.py
+++ b/src/phonometry/building/measurement/uncertainty.py
@@ -621,7 +621,7 @@ def _warn_renamed(old: str, new: str) -> None:
     warnings.warn(
         f"phonometry {old} (ISO 12999-1) is deprecated since phonometry 3.1 "
         f"and will be removed in 4.0; use {new}. For the GUM function use "
-        f"phonometry.uncertainty.{old}.",
+        f"phonometry.metrology.uncertainty.{old}.",
         DeprecationWarning,
         stacklevel=3,
     )
diff --git a/src/phonometry/environment/propagation/air_absorption.py b/src/phonometry/environment/propagation/air_absorption.py
index e3ee8a81f..42b63cfde 100644
--- a/src/phonometry/environment/propagation/air_absorption.py
+++ b/src/phonometry/environment/propagation/air_absorption.py
@@ -59,7 +59,7 @@
 ``exact_midband=True`` to snap the requested frequencies onto that grid and
 reproduce Table 1 exactly.
 
-This module closes the loop with :mod:`phonometry.materials.absorbers.sound_absorption` (ISO 354),
+This module closes the loop with :mod:`~phonometry.materials.absorbers.sound_absorption` (ISO 354),
 whose air power-attenuation coefficient ``m`` (1/m) is defined only through
 the ISO 9613-1 ``alpha`` via :math:`m = \alpha / (10 \log_{10} e)`.
 :func:`air_attenuation_m` returns that ``m`` directly.
@@ -223,8 +223,8 @@ def air_attenuation(
         ISO 354:2003 defers its air power-attenuation coefficient ``m`` (1/m)
         entirely to this ``alpha`` via :math:`m = \alpha / (10 \log_{10} e)`. Use
         :func:`air_attenuation_m` to obtain that ``m`` for
-        :func:`phonometry.sound_absorption.absorption_area` /
-        :func:`~phonometry.sound_absorption.absorption_coefficient`.
+        :func:`~phonometry.materials.absorbers.sound_absorption.absorption_area` /
+        :func:`~phonometry.materials.absorbers.sound_absorption.absorption_coefficient`.
     """
     freqs = np.asarray(frequencies, dtype=np.float64)
     _validate(freqs, temperature, relative_humidity, pressure)
@@ -266,10 +266,10 @@ def air_attenuation_m(
     Convenience composition of :func:`air_attenuation` (ISO 9613-1 ``alpha`` in
     dB/m) with the ISO 354:2003 (8.1.2.1) conversion
     :math:`m = \alpha / (10 \log_{10} e)`
-    (via :func:`phonometry.sound_absorption.attenuation_from_alpha`). It lets an
+    (via :func:`~phonometry.materials.absorbers.sound_absorption.attenuation_from_alpha`). It lets an
     ISO 354 caller feed real atmospheric conditions into
-    :func:`~phonometry.sound_absorption.absorption_area` /
-    :func:`~phonometry.sound_absorption.absorption_coefficient` instead of
+    :func:`~phonometry.materials.absorbers.sound_absorption.absorption_area` /
+    :func:`~phonometry.materials.absorbers.sound_absorption.absorption_coefficient` instead of
     hand-entering ``m``.
 
     :param frequencies: Frequency or frequencies ``f``, in hertz (array-like).
diff --git a/src/phonometry/environment/propagation/outdoor_propagation.py b/src/phonometry/environment/propagation/outdoor_propagation.py
index 5756e9a8c..5f5cc9720 100644
--- a/src/phonometry/environment/propagation/outdoor_propagation.py
+++ b/src/phonometry/environment/propagation/outdoor_propagation.py
@@ -25,7 +25,7 @@
 
 * ``Adiv`` geometrical divergence, :math:`20 \log_{10}(d/d_0) + 11` (Eq. (7));
 * ``Aatm`` atmospheric absorption, :math:`\alpha d` (Eq. (8)) with ``alpha``
-  the ISO 9613-1 coefficient supplied by :mod:`phonometry.environment.propagation.air_absorption`;
+  the ISO 9613-1 coefficient supplied by :mod:`~phonometry.environment.propagation.air_absorption`;
 * ``Agr`` ground effect, both the general per-region method of 7.3.1 with the
   Table 3 functions ``a'/b'/c'/d'`` (Eq. (9)) and the alternative simplified
   method of 7.3.2 (Eq. (10));
@@ -333,7 +333,7 @@ def atmospheric_absorption(
 
     :math:`A_{atm} = \alpha d` with ``alpha`` the ISO 9613-1 atmospheric
     attenuation
-    coefficient (here in dB/m, from :func:`phonometry.air_absorption.air_attenuation`)
+    coefficient (here in dB/m, from :func:`~phonometry.environment.propagation.air_absorption.air_attenuation`)
     at each octave-band midband frequency. Eq. (8) writes ``alpha`` in dB/km
     with :math:`A_{atm} = \alpha_{\text{dB/km}} \, d / 1000`; the two forms
     are identical.
diff --git a/src/phonometry/environment/sources/wind_turbine.py b/src/phonometry/environment/sources/wind_turbine.py
index eb946c1f3..2628a0c9a 100644
--- a/src/phonometry/environment/sources/wind_turbine.py
+++ b/src/phonometry/environment/sources/wind_turbine.py
@@ -15,10 +15,10 @@
   ``ΔL_a`` that decides whether a tone is audible.
 
 The tonal-audibility formula itself is the ISO 1996-2 Annex C one already in
-:mod:`phonometry.environment.assessment.measurement`; what is specific to IEC 61400-11 is
+:mod:`~phonometry.environment.assessment.measurement`; what is specific to IEC 61400-11 is
 how the tone and masking-noise levels and the (Zwicker) critical band are
 determined from the narrowband spectrum. The rating adjustment ``K_T`` is the
-ISO 1996-2 :func:`~phonometry.environmental_measurement.tonal_adjustment`. The
+ISO 1996-2 :func:`~phonometry.environment.assessment.measurement.tonal_adjustment`. The
 full measurement pipeline (binning, regression to standardised wind speeds,
 uncertainty budgets) is out of scope; these are the underlying closed forms.
 """
@@ -109,7 +109,7 @@ def apparent_sound_power_level(
         asterisk marking and the <= 3 dB not-reported rule, subclause 9.3) is
         out of scope here and must be applied beforehand; note its rule set
         differs from the ISO 1996-2 correction in
-        :func:`phonometry.environment.assessment.measurement.residual_sound_correction`.
+        :func:`~phonometry.environment.assessment.measurement.residual_sound_correction`.
     :param r1: Slant distance ``R1`` to the rotor centre, in m.
     :return: The apparent sound power level ``L_WA``, in dB re 1 pW.
     :raises ValueError: If the inputs are invalid.
diff --git a/src/phonometry/hearing/occupational_exposure.py b/src/phonometry/hearing/occupational_exposure.py
index c2306976e..b933b4e8a 100644
--- a/src/phonometry/hearing/occupational_exposure.py
+++ b/src/phonometry/hearing/occupational_exposure.py
@@ -647,5 +647,5 @@ def __getattr__(name: str) -> Any:
         _warn_renamed("ExposureWarning", "OccupationalExposureWarning")
         return OccupationalExposureWarning
     raise AttributeError(
-        f"module 'phonometry.occupational_exposure' has no attribute {name!r}"
+        f"module 'phonometry.hearing.occupational_exposure' has no attribute {name!r}"
     )
diff --git a/src/phonometry/materials/absorbers/rating.py b/src/phonometry/materials/absorbers/rating.py
index 1eb1a87ef..1d8770290 100644
--- a/src/phonometry/materials/absorbers/rating.py
+++ b/src/phonometry/materials/absorbers/rating.py
@@ -481,7 +481,7 @@ def __getattr__(name: str) -> Any:
         canonical = _RENAMED_CONSTANTS[name]
     except KeyError:
         raise AttributeError(
-            f"module 'phonometry.absorption_rating' has no attribute {name!r}"
+            f"module 'phonometry.materials.absorbers.rating' has no attribute {name!r}"
         ) from None
     _warn_renamed(name, canonical)
     return globals()[canonical]
diff --git a/src/phonometry/materials/diffusers/scattering_diffusion.py b/src/phonometry/materials/diffusers/scattering_diffusion.py
index 95b982fbc..30437391d 100644
--- a/src/phonometry/materials/diffusers/scattering_diffusion.py
+++ b/src/phonometry/materials/diffusers/scattering_diffusion.py
@@ -1187,5 +1187,5 @@ def __getattr__(name: str) -> Any:
         _warn_renamed("BASE_PLATE_BANDS_HZ", "BASE_PLATE_BANDS")
         return BASE_PLATE_BANDS
     raise AttributeError(
-        f"module 'phonometry.scattering_diffusion' has no attribute {name!r}"
+        f"module 'phonometry.materials.diffusers.scattering_diffusion' has no attribute {name!r}"
     )
diff --git a/src/phonometry/materials/surfaces/road_absorption.py b/src/phonometry/materials/surfaces/road_absorption.py
index 4bc672066..e287b8ef5 100644
--- a/src/phonometry/materials/surfaces/road_absorption.py
+++ b/src/phonometry/materials/surfaces/road_absorption.py
@@ -34,7 +34,7 @@
   (measured ``alpha`` below ~0.15). Part 2 does not restate the
   transfer-function / reflection-factor / absorption mathematics; Clauses 4,
   5.7 and 6.6 defer it to ISO 10534-2. The core computation therefore lives in
-  :func:`phonometry.impedance_tube.two_microphone_impedance` and is **not**
+  :func:`~phonometry.materials.absorbers.impedance_tube.two_microphone_impedance` and is **not**
   reimplemented here. This module contributes only the Part-2 tube
   geometry/validity helpers (upper usable frequency, microphone-spacing bounds,
   the 250-1600 Hz one-third-octave working range) and the Annex A internal-loss
@@ -815,7 +815,7 @@ def msa_major_axis(
 # --------------------------------------------------------------------------- #
 # ISO 13472-2 spot method: tube geometry / validity helpers.
 # The core two-microphone DSP is NOT here: see
-# phonometry.impedance_tube.two_microphone_impedance (ISO 10534-2, Clause 7).
+# phonometry.materials.absorbers.impedance_tube.two_microphone_impedance (ISO 10534-2, Clause 7).
 # --------------------------------------------------------------------------- #
 def spot_tube_upper_frequency(
     diameter: float, speed_of_sound: float = DEFAULT_SPEED_OF_SOUND
diff --git a/src/phonometry/psychoacoustics/quality/annoyance.py b/src/phonometry/psychoacoustics/quality/annoyance.py
index bee66006f..a127e3741 100644
--- a/src/phonometry/psychoacoustics/quality/annoyance.py
+++ b/src/phonometry/psychoacoustics/quality/annoyance.py
@@ -173,7 +173,7 @@ def psychoacoustic_annoyance_from_signal(
     with :func:`psychoacoustic_annoyance`: ``N5`` from the ISO 532-1 Zwicker
     time-varying loudness, ``S`` from DIN 45692 sharpness, ``R`` from
     ECMA-418-2 roughness and ``F`` from
-    :func:`~phonometry.fluctuation_strength.fluctuation_strength`.
+    :func:`~phonometry.psychoacoustics.quality.fluctuation_strength.fluctuation_strength`.
 
     .. note::
         This composite mixes model families (Zwicker ``N5``/``S``, Sottek
@@ -185,7 +185,7 @@ def psychoacoustic_annoyance_from_signal(
         ``field`` selects the sound field for the loudness, sharpness and
         roughness front-ends; the fluctuation strength ``F`` is always computed
         in the free field, because the Osses 2016 model has no diffuse-field
-        variant (see :func:`~phonometry.fluctuation_strength.fluctuation_strength`).
+        variant (see :func:`~phonometry.psychoacoustics.quality.fluctuation_strength.fluctuation_strength`).
 
     :param x: Calibrated sound-pressure signal (1-D), in Pa after
         ``calibration_factor``.
diff --git a/src/phonometry/room/image_source.py b/src/phonometry/room/image_source.py
index b559d3e81..7db09a113 100644
--- a/src/phonometry/room/image_source.py
+++ b/src/phonometry/room/image_source.py
@@ -471,7 +471,7 @@ def image_source_rir(
         neper per metre (scalar or per-band); the pressure amplitude of each
         path is scaled by :math:`e^{-m r / 2}` (Kuttruff 4.1). Default 0 (air
         absorption neglected). Obtain a physical ``m`` from
-        :func:`phonometry.air_absorption.air_attenuation_m`.
+        :func:`~phonometry.environment.propagation.air_absorption.air_attenuation_m`.
     :param duration: RIR length, s; default the latest image arrival rounded up
         to the next sample.
     :param frequencies: Optional band centre frequencies, Hz, labelling a
diff --git a/src/phonometry/room/reverberation_prediction.py b/src/phonometry/room/reverberation_prediction.py
index b7fc0d02e..9bde68f88 100644
--- a/src/phonometry/room/reverberation_prediction.py
+++ b/src/phonometry/room/reverberation_prediction.py
@@ -42,7 +42,7 @@
 Air absorption enters every model through the ``air_attenuation`` power
 coefficient ``m`` (in neper per metre) as the additive term :math:`4 m V`;
 obtain a physical ``m`` from temperature and humidity with
-:func:`phonometry.air_absorption.air_attenuation_m`.
+:func:`~phonometry.environment.propagation.air_absorption.air_attenuation_m`.
 
 Each model enforces its own mathematical domain on the absorption
 coefficients. Sabine's linear sum is finite for any non-negative coefficient,
@@ -290,7 +290,7 @@ def sabine_reverberation_time(
         coefficient a scalar or a per-band array in ``[0, 2]``.
     :param air_attenuation: Air power-attenuation coefficient ``m``, in neper
         per metre (scalar or per-band); see
-        :func:`phonometry.air_absorption.air_attenuation_m`. Default ``0``
+        :func:`~phonometry.environment.propagation.air_absorption.air_attenuation_m`. Default ``0``
         (air absorption neglected).
     :param speed_of_sound: Speed of sound ``c0``, m/s (default
         :data:`DEFAULT_SPEED_OF_SOUND`, giving the factor ``0.161``).
diff --git a/src/phonometry/speech/sii.py b/src/phonometry/speech/sii.py
index eec0b37fe..b180e7add 100644
--- a/src/phonometry/speech/sii.py
+++ b/src/phonometry/speech/sii.py
@@ -877,4 +877,4 @@ def __getattr__(name: str) -> Any:
     if name == "BAND_CENTRES":
         _warn_renamed("BAND_CENTRES", "BAND_CENTERS")
         return BAND_CENTERS
-    raise AttributeError(f"module 'phonometry.sii' has no attribute {name!r}")
+    raise AttributeError(f"module 'phonometry.speech.sii' has no attribute {name!r}")
diff --git a/tests/test_deprecated_aliases.py b/tests/test_deprecated_aliases.py
index c2580583b..29e1f56a3 100644
--- a/tests/test_deprecated_aliases.py
+++ b/tests/test_deprecated_aliases.py
@@ -7,8 +7,9 @@
 ``"deprecated"`` sentinel). Every alias must warn with the NEP 23 message
 and delegate to the canonical name.
 
-Two generations coexist, each removed with its own release: the 3.1 renames
-and the 3.2 module moves go in 4.0, the 4.0 taxonomy aliases in 5.0.
+The 3.2 module moves were removed in 4.0, as announced; what is left here is
+the 3.1 function and keyword renames and the 4.0 taxonomy aliases, which go in
+5.0.
 """
 
 from __future__ import annotations
@@ -26,26 +27,6 @@
 FS = 48_000.0
 
 
-# --------------------------------------------------------------------------- #
-# Renamed module: phonometry.loudness -> phonometry.loudness_zwicker
-# --------------------------------------------------------------------------- #
-def test_loudness_module_attribute_access_warns_and_delegates() -> None:
-    import phonometry.loudness  # noqa: F401  (PEP 562 shim; import is silent)
-
-    shim = sys.modules["phonometry.loudness"]
-    target = sys.modules["phonometry.psychoacoustics.loudness.zwicker"]
-    with pytest.warns(DeprecationWarning, match=r"loudness\.zwicker"):
-        cls = shim.ZwickerLoudness
-    assert cls is target.ZwickerLoudness
-    with pytest.warns(DeprecationWarning, match="deprecated since phonometry 3.1"):
-        func = shim.loudness_zwicker
-    assert func is ph.loudness_zwicker
-    # __dir__ delegates (and does not warn).
-    assert "loudness_zwicker_from_spectrum" in dir(shim)
-    with pytest.raises(AttributeError, match="phonometry.loudness"):
-        _ = shim.does_not_exist
-
-
 # --------------------------------------------------------------------------- #
 # Legacy snake_case function aliases
 # --------------------------------------------------------------------------- #
@@ -222,7 +203,7 @@ def test_room_volume_explicit_none_stays_silent() -> None:
 # the package root.
 # --------------------------------------------------------------------------- #
 def test_octave_bands_hz_warns_and_delegates() -> None:
-    from phonometry import absorption_rating
+    from phonometry.materials.absorbers import rating as absorption_rating
 
     with pytest.warns(DeprecationWarning, match="use OCTAVE_BANDS"):
         legacy = ph.OCTAVE_BANDS_HZ
@@ -233,7 +214,7 @@ def test_octave_bands_hz_warns_and_delegates() -> None:
 
 
 def test_third_octave_bands_hz_warns_and_delegates() -> None:
-    from phonometry import absorption_rating
+    from phonometry.materials.absorbers import rating as absorption_rating
 
     with pytest.warns(DeprecationWarning, match="use THIRD_OCTAVE_BANDS"):
         legacy = ph.THIRD_OCTAVE_BANDS_HZ
@@ -244,7 +225,7 @@ def test_third_octave_bands_hz_warns_and_delegates() -> None:
 
 
 def test_base_plate_bands_hz_warns_and_delegates() -> None:
-    from phonometry import scattering_diffusion
+    from phonometry.materials.diffusers import scattering_diffusion
 
     with pytest.warns(DeprecationWarning, match="use BASE_PLATE_BANDS"):
         legacy = ph.BASE_PLATE_BANDS_HZ
@@ -255,7 +236,7 @@ def test_base_plate_bands_hz_warns_and_delegates() -> None:
 
 
 def test_band_centres_warns_and_delegates() -> None:
-    from phonometry import sii
+    from phonometry.speech import sii
 
     with pytest.warns(DeprecationWarning, match="use BAND_CENTERS"):
         legacy = sii.BAND_CENTRES
@@ -263,7 +244,7 @@ def test_band_centres_warns_and_delegates() -> None:
 
 
 def test_exposure_warning_warns_and_delegates() -> None:
-    from phonometry import occupational_exposure
+    from phonometry.hearing import occupational_exposure
 
     with pytest.warns(DeprecationWarning, match="use OccupationalExposureWarning"):
         legacy = ph.ExposureWarning
@@ -275,124 +256,56 @@ def test_exposure_warning_warns_and_delegates() -> None:
 
 
 def test_renamed_attribute_shims_reject_unknown_names() -> None:
-    from phonometry import absorption_rating, occupational_exposure
+    from phonometry.hearing import occupational_exposure
+    from phonometry.materials.absorbers import rating as absorption_rating
 
     with pytest.raises(AttributeError, match="phonometry"):
         _ = ph.does_not_exist
-    with pytest.raises(AttributeError, match="absorption_rating"):
+    with pytest.raises(AttributeError, match="absorbers.rating"):
         _ = absorption_rating.does_not_exist
     with pytest.raises(AttributeError, match="occupational_exposure"):
         _ = occupational_exposure.does_not_exist
 
 
 # --------------------------------------------------------------------------- #
-# 3.2 package reorganization: every pre-move public module path must remain
-# importable (silently) for one deprecation cycle. Frozen snapshot; do NOT
-# regenerate from the live tree (that would defeat its purpose).
+# 3.2 package reorganization: its flat module paths were removed in 4.0, as
+# announced. A sample of them, frozen; the point is that they stay gone, and
+# that a stale file left on disk would be caught rather than silently served.
 # --------------------------------------------------------------------------- #
-_PRE_MOVE_MODULE_PATHS = [
-    "phonometry.absorption_rating",
-    "phonometry.absorption_uncertainty",
-    "phonometry.air_absorption",
-    "phonometry.aircraft_atmospheric_absorption",
-    "phonometry.aircraft_noise",
-    "phonometry.airflow_resistance",
-    "phonometry.airport_noise",
-    "phonometry.building_prediction",
-    "phonometry.building_uncertainty",
-    "phonometry.calibration",
-    "phonometry.compliance",
+_REMOVED_FLAT_MODULE_PATHS = [
     "phonometry.core",
-    "phonometry.distortion",
-    "phonometry.dynamic_stiffness",
-    "phonometry.enclosed_space_absorption",
-    "phonometry.environmental",
-    "phonometry.environmental_measurement",
-    "phonometry.facade_prediction",
-    "phonometry.filter_design",
-    "phonometry.flanking_transmission",
-    "phonometry.floor_covering_improvement",
-    "phonometry.fluctuation_strength",
-    "phonometry.frequencies",
-    "phonometry.frequency_response",
-    "phonometry.hearing",
-    "phonometry.human_vibration",
-    "phonometry.impedance_tube",
-    "phonometry.impulse_prominence",
-    "phonometry.installed_structure_borne",
     "phonometry.insulation",
-    "phonometry.intensity",
-    "phonometry.intensity_insulation",
-    "phonometry.lab_insulation",
     "phonometry.levels",
     "phonometry.loudness",
-    "phonometry.loudness_contours",
-    "phonometry.loudness_ecma",
-    "phonometry.loudness_moore_glasberg",
-    "phonometry.loudness_moore_glasberg_time",
-    "phonometry.loudness_zwicker",
-    "phonometry.mechanical_mobility",
-    "phonometry.multiple_shock_vibration",
-    "phonometry.noise_induced_hearing_loss",
-    "phonometry.numerical_propagation",
-    "phonometry.occupational_exposure",
-    "phonometry.ocean_ambient_noise",
-    "phonometry.open_plan",
-    "phonometry.outdoor_propagation",
-    "phonometry.parametric_filters",
-    "phonometry.pile_driving_noise",
-    "phonometry._plotting",
-    "phonometry.psychoacoustic_annoyance",
-    "phonometry.reverberation_prediction",
-    "phonometry.road_absorption",
-    "phonometry.room_acoustics",
     "phonometry.room_ir",
-    "phonometry.room_noise",
-    "phonometry.rotorcraft_noise",
-    "phonometry.roughness_ecma",
-    "phonometry.scattering_diffusion",
-    "phonometry.seabed_reflection",
-    "phonometry.sharpness",
-    "phonometry.ship_radiated_noise",
-    "phonometry.ship_traffic_noise",
-    "phonometry.sii",
-    "phonometry.sonar_equation",
-    "phonometry.sound_absorption",
-    "phonometry.sound_power",
-    "phonometry.sound_power_intensity",
-    "phonometry.sound_power_reverberation",
-    "phonometry.sti",
-    "phonometry.structure_borne_power",
-    "phonometry.survey_insulation",
-    "phonometry.tonality",
-    "phonometry.tonality_ecma",
-    "phonometry.tone_audibility",
-    "phonometry.transfer_stiffness",
-    "phonometry.uncertainty",
     "phonometry.underwater_acoustics",
-    "phonometry.underwater_propagation",
-    "phonometry.underwater_sound_speed",
     "phonometry.utils",
-    "phonometry.vibration_sound_power",
-    "phonometry._warnings",
-    "phonometry.wind_turbine_noise",
 ]
 
 
-@pytest.mark.parametrize("path", _PRE_MOVE_MODULE_PATHS)
-def test_pre_move_module_path_still_imports(path: str) -> None:
+def test_pre_split_module_alias_rejects_unknown_names() -> None:
+    """A 4.0 alias serves what its target has, and nothing else.
+
+    The alias delegates by ``getattr`` on the relocated module, so a typo has
+    to come back as an ``AttributeError`` naming the path the caller wrote,
+    not as a silent ``None`` or a confusing message about the new module.
+    """
     import importlib
-    import warnings
 
-    with warnings.catch_warnings():
-        warnings.simplefilter("error", DeprecationWarning)
-        module = importlib.import_module(path)  # import itself must be silent
-    assert module is sys.modules[path]
-    # A shim that imports but exposes nothing is as broken as an ImportError:
-    # every pre-move path must still surface its public names. ``dir()`` is
-    # silent on the PEP 562 shims, so this stays warning-free.
-    public = [name for name in dir(module) if not name.startswith("_")]
-    assert public, f"{path} imports but exposes no public names"
+    shim = importlib.import_module("phonometry.metrology.levels")
+    with pytest.raises(AttributeError, match="phonometry.metrology.levels"):
+        _ = shim.does_not_exist
+
+
+@pytest.mark.parametrize("path", _REMOVED_FLAT_MODULE_PATHS)
+def test_removed_flat_module_path_raises(path: str) -> None:
+    import importlib
+
+    with pytest.raises(ModuleNotFoundError):
+        importlib.import_module(path)
+    assert path not in sys.modules
+    # The names themselves never moved: the flat API is what they were for.
+    assert hasattr(ph, "leq")
 
 
 # --------------------------------------------------------------------------- #
@@ -535,7 +448,7 @@ def test_the_migration_table_names_real_aliases() -> None:
     import importlib
     import re
 
-    from phonometry._compat import _MOVED_3X, _MOVED_4X
+    from phonometry._compat import _MOVED_4X
 
     table = (
         pathlib.Path(__file__).resolve().parent.parent
@@ -554,10 +467,10 @@ def test_the_migration_table_names_real_aliases() -> None:
         f"no example row for {sorted(moved_from - documented)}"
     )
     for old, new, removed_in in rows:
-        generation = _MOVED_3X if removed_in == "4.0" else _MOVED_4X
-        assert old in generation, f"{old} is not a deprecated path of {removed_in}"
-        assert generation[old] == new, (
-            f"{old} resolves to {generation[old]}, not to {new}"
+        assert removed_in == "5.0", f"{old} names a removal that already happened"
+        assert old in _MOVED_4X, f"{old} is not a deprecated path"
+        assert _MOVED_4X[old] == new, (
+            f"{old} resolves to {_MOVED_4X[old]}, not to {new}"
         )
         importlib.import_module(new)
 
@@ -797,9 +710,9 @@ def test_plotting_shim_re_exports_every_renderer() -> None:
 def test_moved_module_shims_warn_and_delegate() -> None:
     import importlib
 
-    from phonometry._compat import _MOVED_3X, _MOVED_4X
+    from phonometry._compat import _MOVED_4X
 
-    for old, new in {**_MOVED_3X, **_MOVED_4X}.items():
+    for old, new in _MOVED_4X.items():
         shim = importlib.import_module(old)
         target = importlib.import_module(new)
         public = [n for n in dir(target) if not n.startswith("_")]
diff --git a/tests/test_utils.py b/tests/test_utils.py
index 872d7756a..f5bea124b 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -4,7 +4,8 @@
 """
 
 import numpy as np
-from phonometry.utils import _resample_to_length, _typesignal
+
+from phonometry._internal.utils import _resample_to_length, _typesignal
 
 
 def test_typesignal_converts_int_arrays_to_float64() -> None: