The Five Types of Intermittent Mandatory Ventilation
Educational reference — a way to understand and name what a ventilator is doing when it lets a patient breathe between mandatory breaths. It is not a bedside directive, a weaning protocol, or a substitute for a device’s operator’s manual or for clinical judgment.
One word carries the whole idea: permitted
In the taxonomy of ventilator modes, the second axis — the breath sequence — asks how mandatory and spontaneous breaths coexist, and it has exactly three answers: CMV (no spontaneous breaths between mandatory ones), CSV (all breaths spontaneous), and the one this chapter is about, IMV. IMV is formally defined as the breath sequence in which spontaneous breaths (patient triggered and patient cycled) are permitted between mandatory breaths (machine triggered or machine cycled) [established] [Chatburn & Liu 2022].
The load-bearing word is “permitted.” It is not “occur,” not “happen,” not “are present” — it is permitted. That precision exists to head off a seductive misclassification. On a ventilator with an active exhalation valve, a patient can draw a spontaneous breath during a pressure-control mandatory breath. It is tempting to conclude that the mode has therefore become IMV. It has not. If that spontaneous effort lands inside a PC-CMV breath, it does not convert CMV to IMV, because it does not change the mandatory breath rate — the machine is still delivering its set number of mandatory breaths per minute [Chatburn & Liu 2022]. IMV is defined by spontaneous breaths being permitted in the gaps between mandatory breaths, altering how many mandatory breaths the patient actually receives. A spontaneous breath riding on top of a mandatory one is a feature of the exhalation valve, not a change in breath sequence.
Hold onto that, because it is the seed of everything that follows: the entire five-type story is a story about the gaps — what the ventilator is allowed to do with the mandatory breaths that live in those gaps, and under what conditions it suppresses them. Understanding these different forms of IMV is described as essential to recognizing the similarities and differences among many dozens of modes, which matters for clinical application, education, and research [Chatburn & Liu 2022].
SIMV, the synchronization window, and the disappearing “S”
Before the five types, one more foundational distinction, because it explains why the set rate means opposite things in CMV versus IMV.
Historically, plain IMV delivered its mandatory breaths on a rigid clock — a mandatory breath every n seconds no matter what the patient was doing, which could stack a machine breath directly on top of a patient’s own inspiration. SIMV — synchronized IMV — fixed this with a synchronization window: a short period (for example, ~5 s) at the end of the preset expiratory time during which a patient effort is allowed to trigger and synchronize the scheduled mandatory breath. When the patient triggers within that window, the ventilator then stretches the expiratory time so that the set mandatory rate stays roughly constant [established] [Chatburn & Liu 2022; Chatburn 2026 · Five IMV Types].
The rest of the expiratory period (minus a brief refractory interval) is the trigger window. A sufficient patient effort there does something entirely different: it starts a patient-triggered, patient-cycled spontaneous breath, not a mandatory one [Chatburn & Liu 2022].
This trigger-window-versus-synchronization-window split is not bookkeeping trivia. It is precisely what makes the set mandatory rate a maximum for IMV and a minimum for CMV [established] [Chatburn & Liu 2022]:
- In CMV, every valid patient trigger produces a machine-cycled (mandatory) breath, so the patient can only ever raise the total rate above the set value. The set rate is a floor.
- In IMV, efforts that fall in the trigger window become spontaneous breaths instead of mandatory ones, so the number of mandatory breaths delivered can only be the set value or fewer. The set rate is a ceiling.
The taxonomy then makes a deliberate move: it drops the “S.” Nearly every modern ventilator can be patient-triggered, so “synchronized” adds no classificatory information — it does not help you distinguish one mode from another [proposed] [Chatburn 2026 · Five IMV Types]. This is a naming decision, not a clinical dismissal: synchronization remains clinically desirable, and is associated with lower airway pressures and better patient–ventilator interaction [Chatburn 2026 · Five IMV Types]. So in the tags you will see below, “SIMV” on a ventilator’s front panel is classified simply as some form of IMV.
From four “varieties” (2022) to five “types” (2026)
flowchart TD M["IMV = spontaneous breaths<br/>permitted between mandatory"] --> Q1{"Can spontaneous activity<br/>suppress mandatory breaths?"} Q1 -->|no| I1["IMV(1)<br/>mandatory always delivered"] Q1 -->|yes| Q2{"Suppressed by what?"} Q2 --> I2["IMV(2)<br/>spontaneous frequency<br/>above a critical rate"] Q2 --> I3["IMV(3)<br/>a ventilation metric<br/>(e.g. minute ventilation)"] Q2 --> I4["IMV(4)<br/>one big effort flips a breath<br/>from volume to flow cycling"] Q2 --> I5["IMV(5)<br/>flow-cycle after Ti-min<br/>in pressure support"]
Figure 1 — distinguishing the five IMV mechanisms by how (and whether) spontaneous activity suppresses mandatory breaths. Adapted from [Chatburn 2026 · Five IMV Types].
IMV has a long history — it began in the early 1970s as a literal circuit modification (an anesthesia bag, one-way valves, and a continuous fresh-gas flow bolted onto a ventilator so a patient could breathe spontaneously between mandatory breaths, as a weaning aid) [Chatburn 2026 · Five IMV Types]. Over the following half-century it accreted automation, and the corpus has tracked that accretion twice.
The 2022 evolution paper enumerated four distinct varieties of IMV, each adding a layer of automation over the last, moving from manual weaning toward progressively more automatic suppression of mandatory breaths [Chatburn & Liu 2022]. The 2026 five-types paper enumerated five distinct types, IMV(1)–IMV(5) [Chatburn 2026 · Five IMV Types]. That looks like a contradiction, and it was flagged as one — but the full text resolves it cleanly:
- The added type is IMV(5), the Ti-min mechanism found on pressure-support and home/noninvasive ventilators [Chatburn 2026 · Five IMV Types].
- IMV(1)–IMV(4) carry over from the 2022 four-variety scheme [Chatburn 2026 · Five IMV Types].
- Two of the carried-over types are redefined / future-proofed in 2026: IMV(2) has its threshold generalized from fmand to fcrit (so it can absorb a mode like Automode), and IMV(3) has its threshold generalized from a minute-ventilation target to any global ventilatory-performance metric [proposed] [Chatburn 2026 · Five IMV Types].
A framing note the 2026 paper is insistent about: the types are defined by how a mode functions from an engineering-design perspective, kept deliberately independent of how the mode is used clinically. The clinical advantages and disadvantages are then deduced from those design characteristics rather than observed in trials [Chatburn 2026 · Five IMV Types]. This is why the whole scheme is a [proposed] engineering classification — evolving, and explicitly not universally adopted manufacturer nomenclature. The authors note plainly that manufacturers do not use a mode taxonomy at all; the paper’s job is partly to teach clinicians how to decode operator’s manuals [Chatburn 2026 · Five IMV Types].
The five types are best read as answers to one escalating question: under what condition, if any, does the ventilator withhold a mandatory breath?
IMV(1) — the mandatory rate is always delivered
Mechanism. The simplest type. Mandatory breaths are delivered at the preset rate regardless of patient status; spontaneous breaths are permitted in the gaps between them (and those spontaneous breaths may themselves be assisted, e.g. with pressure support, or not). If the patient happens to trigger a mandatory breath, that breath’s expiratory time is nudged slightly so the set mandatory rate stays roughly constant [Chatburn 2026 · Five IMV Types]. Nothing the patient does suppresses a mandatory breath. This is what “SIMV” means on nearly every ventilator.
Advantages. An assured minimum mandatory rate, and — importantly with a caveat — an assured minute ventilation only if the mandatory breaths are volume control. If the mandatory breaths are pressure control, tidal volume (and therefore minute ventilation) is not guaranteed [Chatburn 2026 · Five IMV Types].
Disadvantages. The greatest risk of trigger/cycle asynchrony of the five types, because an imposed mandatory rate and a fixed mandatory inspiratory time are forced on the patient whether or not they fit the patient’s own timing. Historically, this rigidity has a poor liberation record: data suggest IMV(1) prolongs weaning compared with daily spontaneous-breathing trials followed by abrupt discontinuation [Chatburn 2026 · Five IMV Types]. (Landmark 1990s trials found rate-controlled IMV the poorest weaning approach, largely because the rate had to be turned down by hand [Chatburn & Liu 2022].)
Example modes / ventilators. “SIMV” on nearly all ventilators is typically IMV(1); named examples include SIMV Pressure Control on the Medtronic PB 980 and SIMV pressure control on the Nihon Kohden NKV-550 [Chatburn 2026 · Five IMV Types].
IMV(2) — mandatory breaths suppressed when the patient breathes fast enough
Mechanism. The first type that withholds mandatory breaths. If the patient’s spontaneous breath frequency climbs high enough, the mandatory breaths are suppressed, and the tracing then looks like CSV / pressure support. In the classic implementation the ventilator estimates the spontaneous frequency from the interval between the last mandatory breath and the next patient-triggered breath; if that interval is shorter than the mandatory breath period, the patient is breathing faster than the set rate and the mandatory breaths drop out [Chatburn 2026 · Five IMV Types]. It is, in effect, an automatic weaning switch: breathe adequately on your own and the machine steps back; slow down and the mandatory breaths return as a backup.
The fmand → fcrit redefinition. The original definition suppressed mandatory breaths when fspont exceeded fmand (the set mandatory rate). But the mode Automode (Getinge) uses an “apnea time limit” — computed from the set rate and the count of spontaneous breaths since the last mandatory breath — that can make its true suppression threshold lower than the set mandatory rate. To let S/T and Automode share one classification, the 2026 paper replaces “fspont > fmand” with “fspont > fcrit,” where fcrit is either the set mandatory rate or is derived from other settings. This both unifies the current implementations and future-proofs the definition [proposed] [Chatburn 2026 · Five IMV Types].
Advantages. Decreased risk of trigger/cycle discordance, because the mode stops imposing a mandatory rate and inspiratory time once the patient is breathing adequately [Chatburn 2026 · Five IMV Types].
Disadvantages. Minute ventilation is not guaranteed. A patient breathing shallowly or with a rapid-shallow pattern can meet the frequency threshold while still under-ventilating, raising the risk of alveolar hypoventilation and under-assistance [Chatburn 2026 · Five IMV Types].
Example modes / ventilators. Spontaneous/Timed (S/T) on home-care ventilators (e.g., Philips Trilogy); A/C pressure with flow termination ON (ResMed Stellar 150; React Health VOCSN); the Vyaire Avea flow-cycle option; the “Psync” option on the Hamilton C6’s PSIMV+ mode (which is otherwise IMV(1)); and Automode (Getinge). A subtle case worth remembering: on the Zoll bellavista, even “PSV” classifies as PC-IMV(2) rather than PC-CSV [Chatburn 2026 · Five IMV Types].
IMV(3) — mandatory breaths suppressed when a performance metric is met
Mechanism. IMV(3) is the direct answer to IMV(2)‘s hypoventilation problem. Instead of suppressing mandatory breaths on frequency alone, it suppresses them only when a measure of ventilatory performance is met — currently, when total (spontaneous + mandatory) minute ventilation meets or exceeds a preset mandatory minute ventilation. And rather than abruptly cutting all mandatory breaths, it does so gracefully: as spontaneous minute ventilation rises, the ventilator progressively reduces the mandatory rate until mandatory breaths are no longer needed [Chatburn 2026 · Five IMV Types].
The minute-ventilation → global-metric redefinition. The 2026 paper generalizes the definition: mandatory breaths are suppressed when some global ventilatory-performance metric threshold is met — minute ventilation today, but potentially ventilatory drive (via EAdi or Pmus) in the future, so the ventilator could reintroduce mandatory breaths when drive is inadequate, excessive, or erratic [proposed] [Chatburn 2026 · Five IMV Types].
Advantages. Decreased risk of trigger/cycle discordance, like IMV(2), but with a smoother handoff [Chatburn 2026 · Five IMV Types].
Disadvantages. Under-assistance and alveolar hypoventilation are not fully prevented, because the target is total minute ventilation, not alveolar minute ventilation — a set minute volume can still be met by rapid, shallow breathing with a high dead-space fraction [Chatburn 2026 · Five IMV Types; Chatburn & Liu 2022].
Example modes / ventilators. Mandatory Minute Volume / Mandatory Minute Ventilation (MMV) on the Dräger Evita V800; and the more advanced minute-ventilation-driven modes Adaptive Support Ventilation (ASV) on the Hamilton C1 and IntelliVent-ASV, the latter using volumetric capnography to auto-set minute ventilation toward a PaCO₂ target [Chatburn 2026 · Five IMV Types]. IntelliVent-ASV — classified as PC, IMV(3) with both optimal and intelligent targeting — is cited across the corpus as the most technologically advanced current mode [Chatburn & Liu 2022].
IMV(4) — one big effort turns a single mandatory breath spontaneous
Mechanism. The first three types suppress mandatory breaths by the batch, based on rate or minute ventilation. IMV(4) works one breath at a time. Start in volume control: the ventilator has set a fixed tidal volume and flow, which can fight a patient making a large inspiratory effort. From the equation of motion (see Reading Ventilator Waveforms), as the patient’s muscle pressure (Pmus) rises, airway pressure (Pvent) falls by an equal amount. The ventilator watches Pvent, and if it drops past a threshold (for example, 3 cm H₂O), it delivers as much flow and volume as the patient demands. If the effort is large enough, that inspiration switches from volume/time cycling to flow cycling — a form of dual targeting in which volume control converts to pressure control, tidal volume exceeds the set value, and, because the breath is now patient-triggered and patient-cycled, it becomes spontaneous [Chatburn 2026 · Five IMV Types]. That single mandatory breath has, in effect, been suppressed by the patient’s own effort.
Advantages. Mitigates harmful work-shifting, with decreased risk of multiple triggering and better flow synchronization when inspiratory efforts are large [Chatburn 2026 · Five IMV Types].
Disadvantages. Increased risk of tidal-volume over-dosage, precisely because the safety of a fixed set volume is surrendered when volume control switches to pressure control [Chatburn 2026 · Five IMV Types].
Example modes / ventilators. Volume Control with Flow Adaptation on the Getinge Servo-U [Chatburn 2026 · Five IMV Types]. (The 2022 paper also described a second form of IMV(4) — a pressure A/C breath with flow cycling that becomes spontaneous when patient-triggered, e.g. on the Vyaire Avea, with the trade-off of decreased mean airway pressure and possible oxygenation problems [Chatburn & Liu 2022].)
IMV(5) — the new type: Ti-min guards a spontaneous breath from cycling too early
Mechanism. IMV(5) is the type that took the count from four to five, and it lives on pressure-support and home/noninvasive ventilators. It adds a minimum inspiratory time (Ti-min) setting (range roughly 0–4.0 s; as low as 0.1 s on some devices). Every breath is scheduled to be time-cycled and therefore mandatory — but that time cycling is over-ridden if inspiratory flow is above the flow-cycle threshold once Ti-min has elapsed, which makes the breath flow-cycled and spontaneous. Put the other way around: the mandatory breath is suppressed if the patient’s neural inspiratory time is longer than Ti-min; if neural Ti is shorter than Ti-min, the machine holds inspiration for at least Ti-min, guaranteeing a minimum inspiratory duration [Chatburn 2026 · Five IMV Types].
What problem it solves. It fixes a specific pressure-support failure mode: weak efforts causing early flow cycling and hypoventilation of assisted spontaneous breaths — most useful in restrictive disease with premature cycling, and usually unnecessary in obstructive disease [Chatburn 2026 · Five IMV Types]. It should not be confused with the legacy Ti-max safety backup, which time-cycles a pressure-support breath if flow cannot decay (say, because of a mask leak). The taxonomy ignores Ti-max because it is not normal PS function, whereas Ti-min was designed to act as needed during normal ventilation [Chatburn 2026 · Five IMV Types].
Advantages. Decreased risk of hypoventilation caused by low effort and early cycling [Chatburn 2026 · Five IMV Types].
Disadvantages. Increased risk of a late cycle [Chatburn 2026 · Five IMV Types].
Example modes / ventilators. Pressure Support with Ti-min > 0 — the ResMed AirCurve 10 VAuto, Stellar, and Astral, and the Philips Trilogy [Chatburn 2026 · Five IMV Types].
Writing the type into the tag — and how the types combine
IMV(5) introduces a wrinkle the earlier types did not: a single mode can be more than one type at once. Because Ti-min is an over-ride layered on top of whatever else the mode is doing, it combines: S/T with Ti-min > 0 is IMV(2,5), and ResMed’s P-SIMV and V-SIMV (which fix Ti-min at 0.2 s) are IMV(1,5) [Chatburn 2026 · Five IMV Types]. The comma inside the parentheses reads as “both of these mechanisms are active.”
This is possible only because the 2025 taxonomy update gave the IMV type a formal home inside the mode tag. Where the original taxonomy wrote a single, undifferentiated “IMV,” the update writes the type in parentheses inside the TAG — folding the four-to-five IMV evolution into the written mode name [Fajardo-Campoverdi 2025]:
PC-IMV(1)s,s · PC-IMV(2)s,s · VC-IMV(4)d,d · PC-IMV(4)ar,sr
Read one of these left to right and the whole classification unpacks:
control variable (PC or VC) → breath sequence with its IMV type in
parentheses → the targeting letters for the mandatory (primary) and, after the
comma, spontaneous (secondary) breaths. So PC-IMV(1)s,s is pressure-control
IMV of type 1 with set-point targeting on both breath types; VC-IMV(4)d,d is
volume-control IMV of type 4 with dual targeting on both — exactly the mode
family IMV(4) describes. See The Taxonomy of Ventilator Modes for how the
tag is built and what the targeting letters mean.
Two points to keep honest here. First, the update is a labeling refinement, not
a structural change: the 2025 paper restates the ten principles verbatim, keeps
the same three tag components, the same five basic patterns, and the same seven
targeting schemes, and explicitly keeps the English abbreviations “to promote
consistency in scientific publications, databases and medical records”
[Fajardo-Campoverdi 2025]. Second, the same paper’s Table 2 lists IMV(1)–IMV(5)
with the same definitions carried here [Fajardo-Campoverdi 2025]. (The 2025 paper
also adds a second notation — a “variety” in square brackets — but that
concerns non-IMV distinctions like PC-CSVr[Edi] for NAVA, and belongs to the
taxonomy chapter, not this one.)
The horizon: AI-Supervisory IMV
If IMV(1) → IMV(5) is a history of ever-smarter rules for suppressing mandatory breaths, the authors sketch where that trajectory points: a prospective AI-Supervisory IMV [proposed]. In it, an intelligent targeting scheme would continuously screen many signals at once — pressure, flow, volume, computed compliance and resistance, capnography, oximetry, and possibly EAdi or Pmus — and transiently add or suppress mandatory breaths (and modulate the level of inspiratory assistance) whenever a specific physiologic risk is detected: rapid shallow breathing, de-recruitment, too few natural sighs, or a harmful mismatch between effort and ventilation [Chatburn 2026 · Five IMV Types]. A full implementation, the authors suggest, could render the current discrete IMV types — and perhaps discrete modes altogether — obsolete, while serving all three goals of ventilation at once. This is a designed-but-not-yet-built proposal, offered as a direction rather than a product [Chatburn 2026 · Five IMV Types].
Telling the types apart on a tracing
How do you actually identify which type you are looking at? Partly from the operator’s manual and manufacturer classification tables, and partly from the waveforms themselves. The 2026 paper builds the reader’s eye using tracings recorded not from a ventilator display but from a high-fidelity lung simulator (an IngMar ASL 5000). The simulator was chosen deliberately: it avoids manufacturer display inconsistency, superimposes the equation-of-motion variables compactly for timing (pressure in green, volume in blue, flow in black), and — crucially — can plot Pmus, the patient’s inspiratory effort (shown as a red, downward trace), which a real ventilator cannot measure or display [Chatburn 2026 · Five IMV Types]. The reference lung model is an adult with moderate ARDS (resistance 12 cm H₂O/L/s, compliance 39 mL/cm H₂O), usually with just enough effort to trigger so the waveform is not distorted [Chatburn 2026 · Five IMV Types].
The recognition cues, briefly [Chatburn 2026 · Five IMV Types]:
- IMV(1): mandatory breaths appear at the set rate regardless of patient effort — the rate does not budge.
- IMV(2): once the patient breathes fast enough, mandatory breaths vanish and the tracing looks like pressure support / CSV.
- IMV(3): as spontaneous rate rises, mandatory rate falls progressively until it disappears once total minute ventilation clears the threshold (a conceptual trend more than a single-breath sign).
- IMV(4): hard to see on a real display unless you simulate the large effort; look for the switch from volume/time cycling to flow cycling within one breath.
- IMV(5): watch for the time-cycle-to-flow-cycle transition arriving at Ti-min.
A flowchart is provided in the paper for walking through the five types, and the figures mark each breath as circled M (mandatory) or S (spontaneous) [Chatburn 2026 · Five IMV Types]. This is the point where the classification stops being abstract and becomes a reading skill — the same move as in Reading Ventilator Waveforms, where the mode is identified first and the patient is then read off the waveform.
Why go to this trouble? Because the paper’s blunt summary is that manufacturers and even researchers often fail to differentiate the IMV types: manuals are hard to read, the nomenclature is inconsistent, and interfaces vary — so these modes are under-studied and seldom used knowingly, even though some may suit specific goals better than the common alternatives. Worse, failing to recognize a mode as a form of IMV (rather than the CMV the front panel labels it) can quietly bias research, because it changes patient–ventilator interaction, tidal-volume variability, and sedation requirements [Chatburn 2026 · Five IMV Types].
How this chapter connects
- IMV is one axis-2 answer inside the larger three-question framework —
The Taxonomy of Ventilator Modes — which is where the mode tag, the control
variable, and the seven targeting schemes are defined. The
PC-IMV(2)s,snotation only makes sense against that chapter. - Identifying the five types is ultimately a waveform skill — the mode is named first, then the patient is read off the tracing — see Reading Ventilator Waveforms.
- Every type is argued in terms of the three goals of ventilation (safety, comfort, liberation); each type’s advantage/disadvantage pair maps onto one or more of them, and the whole IMV story is framed as serving those goals better than CMV or CSV.
- Learning to recognize and use these modes to mastery — including on the lung simulator that produced the reference waveforms — is the mission of Learning Mechanical Ventilation - The SEVA Program.
- The IMV story does not end in the ICU: a 2024 letter carries this lineage into home and portable ventilators, where the same breath-sequence mechanisms increasingly appear on smaller machines [Chatburn & Hatipoğlu 2024].
Open
- The exact 1:1 verbatim carry-over of the 2022 variety labels onto the 2026 type labels is [open]. The four-to-five evolution is resolved — the added type is IMV(5) (Ti-min), IMV(1)–IMV(4) carry over, and IMV(2)/IMV(3) are redefined/future-proofed — but the 2026 full text does not restate the 2022 four-variety labels word for word, so the one-to-one correspondence should be confirmed against the 2022 paper’s own enumeration rather than assumed [Chatburn 2026 · Five IMV Types; Chatburn & Liu 2022].
- A framing reminder, not a defect. The clinical advantages and disadvantages of each type are largely deduced from engineering design, not established by outcome trials; the corpus is explicit that modes are approved and marketed with little efficacy evidence, and that the mode plays a relatively small role in major outcomes [Chatburn 2026 · Five IMV Types; Chatburn & Liu 2022]. Treat the type-by-type trade-offs as a reasoning framework, not as graded clinical recommendations.
Sources
Every claim above is drawn from the frozen, human-reviewed primary literature in
the machine layer (raw/literature/, one artifact per paper). The short keys used
inline resolve to:
- [Chatburn & Liu 2022] — Chatburn RL, Liu PH. The Evolution of Intermittent Mandatory Ventilation. Respir Care 2023;68(3):417–428. PMID 36195349.
- [Chatburn 2026 · Five IMV Types] — Chatburn RL, Liu PH, Mireles-Cabodevila E. Characteristics of the Five Different Types of Intermittent Mandatory Ventilation. Respir Care 2026 (epub 2026-03-14). PMID 41830556.
- [Fajardo-Campoverdi 2025] — Fajardo-Campoverdi A, Mireles-Cabodevila E, Medina A, Ibarra-Estrada M, Baltazar-Torres J, Chatburn R. Update of the Taxonomy of Mechanical Ventilation Modes. Med Intensiva (English ed.) 2025;49(10):502211. PMID 40480865.
Recent addition — metadata-only:
- [Chatburn & Hatipoğlu 2024] — Chatburn RL, Hatipoğlu U. The Evolution of Intermittent Mandatory Ventilation: Update and Implications for Home Care. Respir Care 2024;69(11):1484–1486. PMID 39379163. Held at title-level thesis only — no abstract or full text is in the frozen artifact, so its specific claims remain
[open]pending a full-text upgrade.
This chapter is the human-layer synthesis of the machine-layer record
Intermittent Mandatory Ventilation and the three raw/literature/ artifacts it
cites (see the sources and derived_from fields).