The Incubator and Gavage at the Paris Maternité: Four Founding Papers, 1883–1898
Four primary texts document the method of premature-infant care developed at the Paris Maternité under Stéphane Tarnier. Below, each is summarized on its own terms in order of publication; a closing section traces how the ideas moved from one to the next, with the points of practical interest to a clinician.
1. Auvard, De la couveuse pour enfants (Archives de tocologie, October 1883)
Written by Alfred Auvard while an interne at the Paris Maternité, this is the first full published account of Tarnier’s incubator in clinical use. It runs in five parts. Part I describes the Maternité couveuses, built to Tarnier’s specifications by Odile Martin: a sawdust-insulated wooden box on a pedestal, a lower compartment holding a ~71-litre hot-water reservoir, a thermosiphon heated by a gas, alcohol, or petroleum lamp, and an upper compartment for the infant’s basket, with air drawn in at the base, warmed as it rises past the reservoir, and vented through the lid. Auvard records that the couveuses were introduced at the Maternité in 1881, that the working temperature, initially 34–35°, had settled to a mean of about 30°, and that variants were in use elsewhere (Pinard at Lariboisière kept 34°; Budin at the Charité ran a gas-heated model with a Regnard regulator and an electric alarm bell). Feeding is by wet nurse for robust infants and by spoon or cup with pure ass’s milk for the weak and very premature; bottles are proscribed; feeds run every two to three hours.
Part II gives clinical and statistical results for 151 infants treated through mid-July 1883, broken out by indication (prematurity and congenital weakness, hypothermic cyanosis and oedema, respiratory distress, resuscitation after apparent death, athrepsia, syphilis, operative deliveries, malformations). Of 93 simply premature infants, 62 survived and 31 died, with intact survivors down to about 1,400 g. Crucially, Auvard supplies a comparative mortality table for infants under 2,000 g: 38% mortality with the couveuse against 66% at Cochin (1882) and 65% at the Maternité before the incubator was installed (1879–81). For hypothermic oedema he sets the couveuse’s 4 deaths in 21 cases against Depaul’s pre-incubator figure of 16 deaths in 20. Part III reports detailed thermometric, pulse, and respiratory tracings, noting that the incubator accelerates pulse and respiration at first while speeding the recovery of body temperature. Part IV compares Winckel’s permanent warm baths — more powerful but harder to supervise — and judges the couveuse preferable for its simplicity. Part V introduces a deliberately cheap model for private practice: a plain wooden box heated by glazed-earthenware hot-water jars (moines), built under Tarnier’s guidance by Galante and obtainable, Auvard writes, from any ordinary carpenter.
2. Tarnier, Des soins à donner aux enfants nés avant terme (Bull. Acad. méd. Paris, séance of 21 July 1885, 14:944–954)
This is Tarnier’s own first public statement, and it takes the form of a demonstration rather than a treatise: he brings two living premature infants before the Académie and narrates their case histories — a surviving twin born 8 June 1885, six weeks in the couveuse (1,020 g falling to 850 g, then recovering to 955 g); her sister (1,105 g, dead at 25 days with no autopsy lesion); and a boy admitted 23 May 1885, 42 days in the apparatus (1,100 → 1,000 → 1,500 g). He pointedly declines to describe the incubator, calling it well enough known to need no description, and offers no statistics.
The substance of the paper is feeding. Tarnier describes gavage as practised at the Maternité: a soft red rubber sonde, the Bailly artificial-nipple cupule, passage of roughly 15 cm to the stomach, the feed run in by gravity, and rapid withdrawal of the tube to prevent regurgitation. He states that gavage has been in regular use at the Maternité since October 1884, after occasional earlier trials. He gives the general feeding rule — meals more frequent the younger and weaker the infant, about 8 g per feed — and warns that overfeeding produces a hypernutrition oedema that, if pushed, ends in fatal gastroenteritis: “là est le danger le plus grand.” He is candid that the best way to dilute the milk is unsettled, and his statement on viability is tentative and personal: he is disposed to believe he can raise infants of six months, and will not despair of going below 180 days. The recorded discussion is valuable in its own right. Blot grants good results but objects that the couveuse is applicable only in hospital and can never pass into private practice, proposing cotton wadding instead; Tarnier replies that it has already passed into private practice, “et pas par mes mains,” and that wadding cannot substitute because the incubator warms the inspired air. Féréol counters with a premature infant saved by two grandmothers and hourly teaspoons of wet-nurse milk; Tarnier concedes such successes existed but were rare. The session closes on a depopulation theme — France’s declining birth rate, and a method that saves seven or eight times as many premature infants.
3. Tarnier & Budin, Accouchement Prématuré Spontané, Traité de l’art des accouchements, vol. 2, chapter XXII (1888)
The treatise chapter is the matured, systematic version of the method, and it is the form in which most later readers encountered it. It supplies what Tarnier’s 1885 communication did not: a full description of the apparatus, comparative mortality figures, and a formalized feeding protocol — while dropping the live-infant demonstration. The incubator description is presented as drawn from Auvard’s 1883 paper. The mortality data are organized into bands (premature mortality falling across roughly 70%, 26.7%, and 9.8% categories; an overall 66% to 36.6%) attributed to Tarnier’s teaching course of 1886, together with a sclérème before-and-after series (181 deaths in 1877–80 against 9 in 1882–85). The chapter gives the installation date as 1880.
On feeding, the gavage apparatus and technique, the feeding rule, and the overfeeding-oedema warning are carried over from the 1885 communication, set in quotation marks as Tarnier’s words. To these the chapter adds a settled dilution protocol for the sterilized milk (cut with water, or with unsalted bouillon) prepared in the double-boiler marmite américaine — the sterilization vessel itself being already in Maternité use, by Mme Henry’s account, from 1883 — together with the named categories gavage mixte and gavage de renfort; the ass’s-milk-and-meat-broth experiment that earlier texts had reported with interest is relegated to a “not yet conclusive” footnote. The viability claim, tentative in 1885, is now stated flatly: clinical viability has come to coincide with legal viability.
4. Henry, Fondation du pavillon des enfants débiles à la Maternité de Paris (Revue mensuelle des maladies de l’enfance, tome XVI, 1898, pp. 142–154)
Written by the sage-femme en chef (head midwife) whom Tarnier recruited to run the incubator work, this is a firsthand account of how the Maternité’s couveuse experiment became an organized hospital service, together with that service’s own outcome data.
On the founding: named sage-femme en chef on 13 July 1881, Henry was shown by Tarnier, in the nurses’ crèche, a couveuse “qu’il venait de faire construire,” and asked to help repeat his earlier experiments. In 1881, she writes, it still had to be proven that an infant could live in the apparatus; the first infant placed in one was nearly moribund and was revived and saved, and others followed. Press attention brought in infants born in town, and with only six couveuses and five nurses the crèche was quickly overwhelmed. To create a dedicated service she founded a subscription society, “l’Œuvre des couveuses” (~10,000 francs), then secured support from Peyron at the Assistance publique and, through Paul Strauss, from the Paris municipal council, which voted the creation in 1891 with 40,000 francs from the pari mutuel. The pavillon, built to Rochet’s plans, opened on 20 July 1893; Henry equipped it with heating apparatus, an autoclave, and fourteen all-glass couveuses built on Tarnier’s model with modifications he specified. She resigned for personal reasons [Ed: simultaneous with Tarnier’s replacement by Budin], left the hospital at the end of June 1895, and closed her statistics on 1 January 1895.
On outcomes: from 20 July 1893 to 1 January 1895 the service received 721 infants — 364 discharged well, 357 died. Setting aside 107 deaths she judged outside fair assessment (24 born before the term of viability, 15 with grave malformations, 68 brought moribund and dead within 24 hours), she reports 250 deaths in 614 infants, a 40.71% mortality. The service deliberately admitted infants below the viable term, both to give parents hope and to test whether infants under six months’ gestation could be raised; one such infant of 5½ months and 780 g lived thirteen days, suckling a little and digesting fairly, before dying in an attack of cyanosis.
On feeding: Henry records that artificial feeding remained dangerous. Of 74 artificially fed infants in poor condition, 31 died (41%), with sterilized cow’s milk giving little advantage (15/38, 39.4%) over feeds where sterilization was not noted (16/36, 44%). She is explicit that sterilizing cow’s milk did not remove its hazards — green diarrhoea, fetid stools, vomiting, distension, persistent erythema — and reports better results from mixed feeding, with milk and unsalted bouillon sterilized together (10–20 g before the breast feed): of 62 infants so fed, 59 were discharged well and 3 died. She also fixes a chronological point used above: Tarnier had them sterilizing milk in the marmite américaine “déjà en 1883,” experimenting with milk uncut, cut with water, or combined with unsalted bouillon. All observations were checked by Tarnier, who compared Henry’s daily watercolour records of stool colour against the infants’ diapers.
On clinical observation: Henry closes with conditions common in débile infants — sclérème, ophthalmia, and respiratory and digestive disorders — including a non-syphilitic nasal-fossa infection that could deform the nose in a manner resembling syphilitic ozena, treated with boric nasal lavage and mentholated borated vaseline. She distinguishes two forms of newborn cyanosis, nervous and mechanical (the latter from gastric or intestinal distension crowding the diaphragm), notes that the nervous form may account for some infants found dead in their cots, and credits Pinard with identifying the ear as a fifth portal of microbial entry alongside the eyes, nose, mouth, and umbilicus.
5. Summary
The four papers track the method’s two pillars, heat and nutrition, as they were assembled, consolidated, and then institutionalized.
The incubator is documented first and most fully by Auvard, not by Tarnier. The apparatus description that later circulates under the Tarnier (and Tarnier–Budin) name is in fact Auvard’s 1883 text, which the Traité chapter borrows by attribution. So the 1888 chapter is a composite: Auvard’s machine joined to Tarnier’s feeding. Two practical consequences follow. First, the canonical “Tarnier incubator” description is Auvard’s, and the often-cited cheap home model — the carpenter’s box heated with earthenware jars — was already designed for private use in 1883, which is what Tarnier meant in 1885 when he told Blot the device had already passed into practice. Second, there is a one-year inconsistency to keep in view: the Traité gives the installation as 1880, yet two independent primary witnesses point to 1881. Auvard, writing contemporaneously in 1883, dates the couveuses’ introduction to 1881; Mme Henry, who was named sage-femme en chef on 13 July 1881 and put in charge of the incubator work, recalls Tarnier showing her a couveuse “qu’il venait de faire construire” on that date and frames 1881 as the year it first had to be proven that an infant could live in the apparatus. Tarnier’s own 1885 note gives no year. The weight of firsthand evidence favours 1881, leaving the Traité‘s 1880 — drawn into the chapter that otherwise quotes Auvard — as the outlier.
The evidence for the incubator is also older than Tarnier’s own voice on it. Tarnier’s 1885 communication offered only anecdote — three cases — and the mortality tables in the 1888 chapter come from his 1886 course. But a controlled comparison was already in print in 1883: Auvard’s 38% versus ~65% mortality for infants under 2,000 g, with and without the couveuse. For a clinician judging provenance, the quantitative case for warmth as treatment predates Tarnier’s published demonstration; the numbers belong to Auvard.
The feeding method is the part that is genuinely Tarnier’s, and it is the one that visibly matured across the texts. In 1883 Auvard fed by spoon and cup with ass’s milk. By 1885 Tarnier had introduced and regularized gavage by soft sonde, with a fixed technique and a clear caution against overfeeding, but admitted the dilution of the milk was unsettled — even though milk sterilization in the marmite américaine was, per Henry, already routine from 1883. By 1888 the dilution protocol was worked out and the gavage variants were named and codified. The same trajectory shows in the rhetoric: a claim Tarnier advanced in 1885 as a personal hope — that the clinically viable infant might reach the legally viable term — is asserted as fact in 1888.
Henry’s contribution is the third dimension: organization, and an unsentimental audit. Her paper is where the apparatus becomes a service — funded, housed, and staffed — and where the method’s limits are stated plainly. The roughly 41% mortality she reports for 1893–95, after honestly defalcating the moribund and the previously unviable, is a useful corrective to the optimism of the treatise: warmth had been largely solved, but nutrition had not. Her artificial-feeding data make the point sharply — sterilizing cow’s milk conferred little survival advantage (39.4% vs 44%), and mixed feeding with breast milk did far better — which aligns with Tarnier’s own 1885 warning that overfeeding, not cold, had become the principal danger for the infant the incubator had kept alive.
The clinical bottom line is that modern premature care here rests on contributions that can be cleanly separated: the incubator and its first outcome data (Auvard, 1883); gavage feeding with its dosing discipline (Tarnier, 1885, codified 1888); and the translation of all of it into a sustained hospital service, with the sobering mortality figures that came with honest accounting (Henry, 1893–95). The treatise that fused the first two is the reason the package is remembered under one or two names; these four papers are how the contributions, the dates, and the limits can be sorted out.
Last Updated on 06/28/26