A planet announced in 1963, revised in 1969

When the Whole Field Moved

In 1969 Peter van de Kamp reported that thirty years of photographs showed Barnard’s star swinging round a 25-year orbit with a semi-axis of 27.5 milliarcseconds, the pull of a companion 1.7 times the mass of Jupiter. Rebuild that orbit from his own yearly table (this page gets 28.9), bring in the independent measurement that found 4 ± 3, and see how far a new lens cell moved the stars on the same telescope in 1949.

The claim, drawn from its own table

Each dot is one year of plates from the Sproul Observatory refractor, 1938 to 1967, as van de Kamp printed them after removing parallax, proper motion and acceleration. Dot area follows his printed weight, except that the lightest one or two years are drawn at a minimum size. The red curve is his orbit (period 25 years, eccentricity 0.75, periastron 1950), refitted here to his numbers. One table unit on the plate is 1.887 milliarcseconds (mas).

printed, 196927.5 mas

refitted here28.9 mas

worst-case rounding bound± 3.4 mas

A reconstruction from his rounded, already-reduced yearly means: on a worst-case rounding bound, the claimed scale is reproduced within the precision of the rounded table. His plate reductions are not replayed.

Computing the orbit from the table…

The claim, at the strength it was printed

Barnard’s star crosses the sky faster than any other star, and it is the nearest single star to the Sun. From 1938 onward the Sproul Observatory at Swarthmore College photographed it year after year with a long refractor, and measured on each glass plate where the star sat among a few background stars. Take out the star’s straight-line motion, its slow curvature, and the yearly nod of parallax, and what remains ought to be nothing but measuring error. Peter van de Kamp found that what remained had a shape.

Twenty-four early plates (1916–1919) and thirty consecutive years of photographic observations of Barnard’s star covering the interval 1938–1967 confirm orbital motion with a period of 25 years and semi-axis major of 0″0275. Assuming a value of 0.15 ⊙ for the mass of Barnard’s star, the mass of the companion is found to be 0.0016 ⊙, or 1.7 times the mass of Jupiter. Peter van de Kamp (1969), Parallax, Proper Motion, Acceleration, and Orbital Motion of Barnard's Star. The Astronomical Journal 74(2), 238-240. doi:10.1086/110799, abstract, p. 238

He printed its precision, and he tested it on a year the solution had not seen. Both belong to the claim at full strength. The precision is a probable error, a different convention from the reported error the control uses below, and this page never combines the two.

The probable errors of the geometric elements as well as of the semi-axis major are below ±0mm0001 or ±0″002. The recently measured yearly mean for 1968, based on 144 plates taken on 36 nights agrees very closely with the ephemeris based on the present 1916–1967 solution, the residuals being 0mm0000 in x and +0mm0003 in y. Peter van de Kamp (1969), Parallax, Proper Motion, Acceleration, and Orbital Motion of Barnard's Star. The Astronomical Journal 74(2), 238-240. doi:10.1086/110799, p. 240

That was the second telling. The first, in 1963, had a 24-year period, a 24.5 mas semi-axis and a companion of 1.6 Jupiter masses. The 1969 paper added five years, remeasured plates from the critical years around periastron, and moved declinations taken on one emulsion onto the system of another. The paper’s solution rests on 3,036 plates, 10,452 exposures and 766 nights, with a total weight of 2,056.

The paper’s Table I prints the yearly means themselves: 30 rows from 1938 to 1967, plus one aggregate of the 24 early plates, which steadies the motion solution over half a century but is not part of the orbital analysis. Each position is a whole number of units of 0.0001 mm on the plate. At 18.87 arcseconds per millimetre that unit is 1.887 mas, and the claimed semi-axis, 1.46 micrometres on the glass, is about fifteen of them.

How this page refits it

Van de Kamp chose the shape of the orbit from 10 normal points, trying periods from 22 to 29 years, eccentricities from 0.60 to 0.90 and periastron dates from 1948.0 to 1952.0, and adopted 25 years, 0.75 and 1950. This page takes those three elements as given. For each coordinate it solves Kepler’s equation at every epoch and fits a constant and two orbit coefficients by weighted least squares with his printed weights; the largest singular value of the resulting two-by-two matrix is the orbit’s semi-axis. Nothing else is free: no drift, no trend, because the table has already had them removed.

Your hands on the fit

Weights his printed weights every year equal

Trial period (the page’s exploration)

period, years	semi-axis, mas	weighted misfit

Equal weights give 27.4 mas; adding the early aggregate gives 28.91. Among these trial periods, with the eccentricity and periastron held at the values he adopted alongside it, the misfit is smallest at the published 25 years, whichever weights are chosen. The other periods are the page’s exploration, not the paper’s, and section III refuses to attach the control’s sensitivity to them.

Checked against his own page, three ways

The headline. The refit gives 28.894 mas against 27.5 printed, a gap of 1.39 mas. Because every printed position was rounded to a whole unit, the fit can only be pinned so tightly: carrying a worst-case half unit through the same least-squares operator bounds the semi-axis to within 3.26 mas, and with the headline’s own last digit that rounds up to 3.4 mas. On that bound, the claimed scale is reproduced within the precision of the rounded table. The bound is a worst case, with every rounding aligned against the fit. Rounding errors that fall independently would move the refit by only about 0.21 mas (one standard deviation), so rounding does not explain the 1.39 mas gap: his solution differs from this refit in ways the printed table does not record. It comes out larger than printed, not smaller, so nothing on this page is working against a weakened claim. It is a reconstruction from already-reduced yearly means, not a replay of his plate reductions.

The four geometric elements. The paper prints the orbit’s coefficients too: A = +2.3, B = +8.9, F = −6.1, G = +11.4, in table units. The refit gets 2.41, 9.32, −5.76 and 11.72, each within its worst-case rounding bound, though not all within what independent rounding would typically allow. Run his printed elements through the same inversion and they give a semi-axis of 28.15 mas and an inclination of 68.8°, which rounds to the 69° he printed. The semi-axis does not round to his: it is 0.65 mas above the printed 27.5, more than rounding of those elements can move it (0.19 mas). We could not find the route from his printed elements to his printed semi-axis. All three numbers sit well inside the tolerance, so the anchor stands either way.

His residuals, regenerated from his elements

Table I also prints what his orbit left over at every epoch. Give the engine only his four printed elements and it regenerates those two columns: 61 of the 62 printed residuals agree within their rounding bound (about 2.1 units; median difference 0.31). Circles are right ascension, squares declination.

The one that does not is 1953.51 in declination. The table prints a remainder of +7 and a residual of +8; the orbit and that residual together imply a remainder near +2.8, a disagreement of 4.2 units against a bound of 2.2. One of the two printed numbers is probably a misprint. The paper’s own Figure 1 can say which: digitised here, with each panel calibrated against the printed values of the other 29 years, it plots that year at +2.0. The same reading of the right-ascension panel, where table and residual agree, returns +14.0 for a printed +14, and the method misses the printed value of any other year by at most 1.4 units. The figure sides with the residual, so the remainder is the likelier misprint. The anchor still keeps the table exactly as printed; had the remainder been +3, the refit would give 29.07 mas instead. The dot is ringed on the chart above, where it sits where the table puts it.

What the table alone prefers

An instrument that jumps can draw a shape like an orbit. So set a rival against the orbit on the same yearly means: a single offset switched on between the telescope’s documented changes at 1941.82 and 1949.21, plus free linear and quadratic drift. The weighted misfit is 76,419 for that rival, 56,176 for the orbit, and 186,118 for no signal at all. On its own table the orbit is the better description, even though the rival has more free numbers per coordinate. (The orbit’s three elements were chosen from these same data by trial, and the rival’s dates were not, so the contest is tilted towards the orbit.) That is the claim at full strength: nothing inside this table gives it away. It took other telescopes.

The numbers the page read from Table I

epoch	Rx	Ry	weight	vx	vy

Units of 0.0001 mm. Transcribed by hand from the ADS scan and read against it twice. Only the fields the page uses are shown; the paper’s table carries more.

Two other telescopes, the same orbit

In 1973 George Gatewood and Heinrich Eichhorn reduced 241 plates of Barnard’s star taken at the Allegheny and Van Vleck observatories, with a reduction method that treats the reference stars differently from the one used at Sproul. They noticed what the chart above shows: most of the claimed swing is in one coordinate, right ascension.

almost all of the perturbation observed by van de Kamp occurs in the x positions, thus we will concern ourselves only with these Assuming the published orbital elements, we find α = 0″004 ± 0″003. The value given by van de Kamp is 0″028 ± 0″003. George Gatewood and Heinrich Eichhorn (1973), An unsuccessful search for a planetary companion of Barnard's star (BD +4°3561). The Astronomical Journal 78(8), 769-776. doi:10.1086/111480, p. 774

They add that the correlation between their positions and van de Kamp’s predicted variation is 0.06, that if we subtract the predicted variation from each observation, we increase, not decrease, the sum of the squares of the residuals, and that when they tried to improve the orbit the elements changed radically on the first step and by the second iteration e had exceeded 1.0, which is no bound orbit at all. Their conclusion: our observations fail to confirm the existance of a planetary companion to Barnard’s star. (The spelling is theirs.)

Their figure, digitised, against the orbit they tested

the 1969 orbit, from its printed elements
that curve scaled by 0.09 to 0.24
their normal points, digitised
their dashed curve, digitised

Their Figure 1 averages their positions in each tenth of the orbit’s phase. The blue dots are those 10 normal points, digitised from the scan (marker size is their weight class). The red curve is the x variation van de Kamp’s printed elements predict, drawn by this page’s engine. The grey ticks are their own dashed curve, digitised: away from periastron it sits on the engine’s curve with a scatter of 0.6 mas over 22 dashes, so the curve they tested is the curve this page draws.

The digitised normal points, as numbers

phase	x, mas	marker	weight class, as printed

Read from the scan and calibrated by this page’s engine; a transcription with its own error, not their data.

The points hug zero where the orbit plunges. As a consistency check, and only that, the page fits one scale factor to the digitised points: how much of van de Kamp’s curve do they contain? Their weights are printed only as ranges for each marker size, so every one of the 1,024 combinations of lowest and highest weight is tried. The answer lies between 0.09 and 0.24 of the claimed orbit, and their own published result, 4 over van de Kamp’s 28 as they quoted it, is 0.14. Their summary and their figure agree. The digitised figure is a transcription with an error budget of about 1 mas per point, not their data.

III

Could that measurement have seen the orbit?

A control that could not have found the effect proves nothing by not finding it. The strongest way to ask is to plant the claimed orbit, at its claimed size, into the control’s own observations and run their unmodified reduction to see whether it comes back. That cannot be done here. The control authors wrote that the 610 individual observations may be obtained in punched card form from the author. They were not published with the paper, and this page found no public copy; the paper is based on Gatewood’s doctoral dissertation, which this page did not check for them. So this page uses the weaker, honest grade: the control authors’ own reported error, set against the claimed size, at the one orbit that error belongs to.

Grade B: published sensitivity, at the published orbit only

claimed semi-axis (1969, printed)27.5 mas
their fit, published orbit assumed4 ± 3 mas
claimed ÷ reported error9.17
(claimed − found) ÷ reported error7.83
found as a share of claimed15 per cent

Computing…

The verdict here is the page’s disclosed rule, not a probability: a claimed size at least 3 reported errors from zero counts as well above the control’s error scale. The actual ratio is 9.2, so any threshold up to that gives the same answer. The error is the control authors’ reported error; van de Kamp printed probable errors, a different convention, and the two are not mixed here. Change the trial period in section I and this panel refuses: their error belongs to the published 25-year orbit and its orientation, not to any other.

What this page will not do

Press either button to see why the page declines.

The verdict, dated

Verdict of the scientific record

ARTEFACT

As of 2026-09-22, for the historical 25-year, Jupiter-scale astrometric orbit. Decided by the instrument: independent telescopes at two other observatories did not recover it, another field on the same telescope showed an instrumental jump at the orbit’s critical date, and the claimant himself then wrote that the long-period signal had been enhanced by that instrumental change. It took 6 years on this page’s convention, from the 1969 revision to the claimant’s 1975 statement; the field did not agree on a single day, and later planetary claims for this star continued for decades. A different method later spoke to the same claims: after Doppler monitoring of the star spanning 25 years, 1987 to 2012, Choi and colleagues wrote in 2013 that Previous claims of planets around the star by van de Kamp are strongly refuted. This does not mean Barnard’s star has no planets. It has at least four, much smaller ones (section VI).

What would change it: an independent, calibrated astrometric record that recovers the historical orbit’s period, phase, size and orientation, with the reference-star and telescope changes treated in a way that survives independent checking. Finding other planets around the star would not.

The claimant’s reply

Van de Kamp did not simply hold his ground. In 1975 he had every plate remeasured on a two-coordinate machine, set aside the plates from 1938 to 1949 because of the instrumental changes, and analysed 1950 to 1974, with the early plates still anchoring the long-term motion. He named the 1949.21 change as the replacement of the old aluminum cell by a new cast-iron cell, accompanied by a change in emulsion from Eastman Kodak C to G, and found that his earlier data fit the new analysis once corrections of +2.5 micrometres in right ascension and −1.0 in declination were applied over 1942 to 1948. Then, of the orbit this page rebuilds:

It appears now that the emergence of the original 24–26-yr perturbation (van de Kamp 1963, 1969a) was spuriously enhanced by the instrumental equation of 1949. Peter van de Kamp (1975), Astrometric study of Barnard's star from plates taken with the Sproul 61-cm refractor. The Astronomical Journal 80(8), 658-661. doi:10.1086/111791, p. 660

He kept a planetary reading: a 11.5-year perturbation with a semi-axis of 9.3 mas, which he called confirmed, and a 22-year one of 5.3 mas, which he called less well determined. Both are far smaller than the 27.5 mas this page rebuilds. For the shorter one he pointed to his critics’ own material: The short-period orbital effect in R.A. also is present in the Allegheny–Van Vleck material (Gatewood 1972), which gave a semi-amplitude in right ascension of +0.39 ± 0.12 micrometres against his own 0.42. He noted that its errors were appreciably larger, and that the close agreement of the values for α1x from the different materials may be worth noting. That is a reply about the shorter signal, not about the orbit the control tested. In his tag above, 1969a is the paper this page rebuilds and 1969b is his August alternative, in the timeline below. The exchange was not a feud on paper: Gatewood and Eichhorn thank van de Kamp for refereeing their first version and forcing them to sharpen their arguments, and van de Kamp’s 1975 paper thanks John L. Hershey for his contribution to the computational analysis.

1963 The first orbit: 24 years, 24.5 mas, 1.6 Jupiter masses.
1969, March The revision this page rebuilds: 25 years, 27.5 mas, 1.7 Jupiter masses.
1969, August His alternate analysis of 1938 to 1968: two companions on circular orbits of 26 and 12 years, 1.1 and 0.8 Jupiter masses. He wrote that on the basis of observations only, it is not possible at present to decide which interpretation is to be preferred.
1973 Hershey finds a jump in another Sproul field at the lens-cell change; Gatewood and Eichhorn find 4 ± 3 mas.
1975 The claimant: the long-period signal was spuriously enhanced by the instrumental equation of 1949.
2013 Doppler monitoring over 25 years; its authors reject van de Kamp’s planet claims (quoted above).
2024 to 2026 Radial velocities find, confirm and model four small planets with periods of days.

What moved in the telescope?

The sophisticated objection to all of this runs: a lens cell is changed, so what? Every star on the plate sees the same lens, the reduction measures the target against its neighbours, and a shift common to all of them cancels. That is true only if the shift is uniform across the plate, or varies linearly across it, which the reduction’s linear plate constants absorb. The answer comes from a different star on the same telescope.

John L. Hershey, also at Sproul, remeasured 423 plates of the field of AC +65°6955 (Gliese 793), taken from 1937 to 1969, on an automated measuring machine, 12 stars in all. Its positions jump.

This coincides with the installation of a new cell for the 24-inch objective lens and a change in photographic emulsion in 1949.2. No discontinuity was found in y for the parallax star or the reference stars. John L. Hershey (1973), Astrometric analysis of the field of AC +65°6955 from plates taken with the Sproul 24-inch refractor. The Astronomical Journal 78(5), 421-425. doi:10.1086/111436, p. 422

This is not Barnard’s star and not its plates; the page claims no more than a shared telescope and a shared date. The date is the point: 1949.2 falls in the year before the claimed periastron, where the orbit does its fastest moving. Hershey’s Table IV gives, for each star, the mean of 10 nights of x residuals after that date minus the mean of 10 before, and each star’s distance from the plate centre. The chart below is those two columns, nothing more. They are residuals from plate reductions that fitted linear plate constants to all eleven reference stars, so each is a shift against the frame those stars define, not an absolute move. Against that frame the parallax star at the centre shifted 3.3 micrometres and the reference stars nearest it shifted less; the outer ones come out negative, as residuals from such a fitted frame partly must.

The field, and how each star shifted against its frame at 1949.2

Stars at their plate positions from Hershey’s Table II; arrows are the Table IV shifts in x against the reference frame, drawn at 2 cm of plate for the target’s 3.3 micrometres, all along one axis because the jump was only in x. Radii recomputed from Table II around the parallax star match Table IV to within 0.3 cm for 10 of the 11 reference stars. Star 7 is the exception: Table II prints for it the same y as star 3, which would put it 2.2 cm out, while Table IV says 3.2; the paper’s own Fig. 2, digitised here, puts it 3.0 cm out, so it is drawn hollow at that position.

Table IV and Table II, as the page read them

star	radius, cm (Table IV)	shift in x, micrometres	x, mm (Table II)	y, mm (Table II)

As printed. Star 7’s Table II y repeats star 3’s and looks misprinted; the plate map draws it at its Fig. 2 position.

A shift that grows towards the centre does not cancel. The target sits in the middle of its reference stars, so a reduction that ties it to them inherits the difference between how much it shifted and how much they shifted on average. That is what “the whole field moved” means here: not by one amount, but by a pattern.

Put the line through the reference stars yourself

Here is the second layer’s question. Hershey added to his plate reduction a term proportional to distance from the centre, fitted with the linear plate constants to the reference stars on every plate. The target is at the centre, so the correction there is an extrapolation. With the correction set to act at the target’s true position, A study of the residuals of the parallax star showed that over-correction was present. The effect would not be expected to peak sharply at the plate center, and after several trials, the value of ρ for the parallax star was adopted as 1.2 cm. He says so openly. The page asks what Table IV alone says about that choice, using the reference stars only, never the target.

Leave out a reference star Target’s assumed radius, from 0 cm (where it sits) to 1.2 cm (Hershey’s adopted value)

Computing…

With all eleven reference stars the line is jump = 2.66 − 0.501 × radius, in micrometres and centimetres. Leave out each star in turn and the slope stays negative every time, 11 of 11, from −0.561 (without star 11) to −0.416 (without star 4). The radial pattern is not one star’s doing.

The target is where the table gets uncomfortable. At its true position, radius 0, the reference line predicts a jump of 2.66 micrometres against the 3.3 it showed, a residual of 0.64. At the adopted 1.2 cm it predicts 2.06, and the residual grows to 1.24. On Table IV’s own numbers, then, the adopted radius makes the reference stars’ prediction of the target’s 1949 jump worse, not better. That need not make Hershey wrong: he chose the value from his full plate-by-plate residuals over three decades, which Table IV does not contain, and the target’s own jump has three published measures (3.5 in his text, 3.3 in Table IV, 2.9 from his step solution, +1.5 minus −1.4). What it does show is that the correction at the centre depends on a number adopted after looking at the target. The shared instrumental pattern is solid; any numerical correction for a central star, Barnard’s included, is a model.

The line here is the page’s simplification: it leaves out the linear plate terms that Hershey’s reduction carried. Refit Table IV with those terms added, positions from Table II measured from the parallax star and star 7 at either of its two printed positions, and the target’s residual becomes 0.09 to 0.23 micrometres at radius 0 and 0.78 to 0.90 at the adopted 1.2 cm. Smaller numbers, the same direction: the adopted radius still makes the prediction worse.

Hershey also warned against reading his field as a law for every plate: Also, there is no apparent discontinuity in some Sproul plate series where the plate coverage is strong and manual measurements would suffice to detect it. Table IV prints no uncertainty for any star, so this page fits no weights and draws no error bars.

The instrument against the planet, on the glass

All in micrometres on the photographic plate, where one micrometre is 18.87 mas. The claimed orbit’s full swing in right ascension, from its printed elements, is 2.33 micrometres (44 mas). Hershey’s target jumped 1.4 times that between the ten nights before and the ten nights after 1949.2; van de Kamp’s own 1975 correction to his 1942 to 1948 plates is 1.07 times it. These are different stars and different fields, set side by side for scale only.

The planets Barnard’s star does have

In 2024 a team using the ESPRESSO spectrograph confirmed a planet on a 3.15-day orbit, pulling the star to and fro at 55 ± 7 cm/s, with a minimum mass of 0.37 ± 0.05 Earth masses, and three more candidate signals. In 2025 independent MAROON-X data, together with ESPRESSO’s, confirmed all four, with periods of 3.154, 4.124, 2.340, 6.739 days and minimum masses from 0.19 to 0.34 Earth masses in their joint fit. A 2026 study of the system’s stability treats the four as confirmed. The 2025 paper sums up what came before:

It has a long history of claimed planet detections from both radial velocities and astrometry. However, none of these claimed detections have so far withstood further scrutiny. Ritvik Basant, Rafael Luque, Jacob L. Bean, Andreas Seifahrt, Madison Brady, Lily L. Zhao, Nina Brown, Tanya Das, Julian Stürmer, David Kasper, Rohan Gupta and Guðmundur Stefánsson (2025), Four Sub-Earth Planets Orbiting Barnard's Star from MAROON-X and ESPRESSO. The Astrophysical Journal Letters 982(1), L1. doi:10.3847/2041-8213/adb8d5, abstract (arXiv:2503.08095v1, CC0)

None of this vindicates the 1969 orbit; these are different planets, found by a different method, at periods of days rather than decades. For scale, the claim’s own numbers are enough: the size of a planet’s astrometric pull on its star grows with the planet’s mass and with the two-thirds power of its period, so at its minimum mass the 2024 planet would move Barnard’s star about 0.09 microarcseconds, roughly 297,000 times less than the claimed 27.5 mas (taking 1.7 Jupiter masses as 540 Earth masses). Radial velocities give only a minimum mass, and the true mass is larger by an unknown factor set by the orbit’s tilt, so that figure is a floor, not a ceiling.

VII

The check

Recomputing every printed figure…

Hashing the frozen files…

Recomputed in your browser, and again by the verifier

The orbit refit from the 30 yearly means (weight 2,032): 28.894 mas, inclination 69.5°, weighted misfit 56,176. The rounding bound, 3.26 mas, is computed from the same least-squares operator, not typed in.
The semi-axis implied by the paper’s printed elements, 28.15 mas (1.492 micrometres against the printed 1.46), and the regeneration of his printed residuals, 61 of 62.
The grade B ratios, 9.17, 7.83 and 0.145, from the control’s published numbers, and the verdict word from the same function: the control could have confirmed it.
The digitised figure’s calibration: axis ticks within 0.38 mas of a straight line, the drawn zero line 0.43 mas from the zero tick, the two points the authors drew twice agreeing to 0.15 mas.
The reference-star line, its eleven omissions, and the target’s residual at both ends of its labelled interval.

Uncertainties the page names

The yearly means are rounded to a whole unit and are the output of reductions this page cannot replay: colour, emulsion, personal-equation, proper-motion and acceleration corrections. The refit is a reconstruction of the orbit’s scale, not of his analysis.
His printed semi-axis does not follow from his printed elements by the standard inversion; one Table I cell (1953.51, declination) disagrees with its own residual, and his digitised Figure 1 sides with the residual. Both figures are transcriptions with their own error, and the page does not edit the table.
The control’s fit is published, not rerun. Its figure is digitised, with weights known only as ranges.
Hershey’s field is another star’s field. Table IV has no per-star errors, and its shifts are residuals against a fitted frame, not absolute moves. The page’s reference line leaves out his linear plate terms; adding them shrinks the numbers and keeps the direction. Star 7’s Table II position looks misprinted.
The anchor’s tolerance is a worst-case rounding bound. Independent rounding would move the refit far less, so the gap between the refit and the printed value is a difference of solution, which the table cannot show.
The historical probable errors (van de Kamp) and the control’s reported errors (Gatewood and Eichhorn) are different conventions and are never combined.

Every free choice

Weights: his printed weights (default) or equal. Trial period: 25 (published) or four others, explicitly the page’s exploration.
The rival model’s form: one offset between the documented 1941.82 and 1949.21 changes, plus linear and quadratic drift.
The grade B rule of 3 reported errors, disclosed above with the largest rule that gives the same answer.
In the figure check: dashes within 0.03 of periastron in phase are left out of the curve comparison, because the draughtsman smoothed the steepest part; the per-point budget is rounded up to 1 mas.
In the residual regeneration: one constant per coordinate, the weighted mean, because the paper does not print its own.
In the second layer: which reference star to leave out, and the target’s assumed radius between its two labelled endpoints.

Not done, and why

No planted-orbit test in the control’s data (grade A): the observations were not published with the paper, and this page found no public copy. The paper is based on Gatewood’s doctoral dissertation, which this page did not check for them.
No false-positive rate for the 1969 method: the plate cadence and the joint error model behind the yearly means are not published, and white noise through this fit would not be his procedure.

What the page adds, and how far that claim goes. We searched the shipped Wasteland corpus, web results for Barnard's star interactive reanalyses, and ADS- and arXiv-indexed results for van de Kamp and Hershey on 2026-09-22 and did not find an interactive joining the 1969 rounded orbit table, the independent published sensitivity, and a reference-only audit of Hershey's adopted target radius. Those were bounded searches, not a survey of everything ever published.

The verifier for this page recomputes every figure above from the same modules and data files, reads this page’s text, and fails on any disagreement. The site publishes it beside the page.

Refitted from its own yearly table, the 1969 orbit comes out at 28.9 mas against 27.5 printed; the independent measurement found 4 ± 3 mas, and on its authors’ own error the claimed orbit sat 9.2 errors above zero.

Sources

Peter van de Kamp (1969), Parallax, Proper Motion, Acceleration, and Orbital Motion of Barnard's Star. The Astronomical Journal 74(2), 238-240. doi:10.1086/110799. doi.org/10.1086/110799. The claim, Table I, the printed elements.
Peter van de Kamp (1969), Alternate Dynamical Analysis of Barnard's Star. The Astronomical Journal 74(6), 757-759. doi:10.1086/110852. doi.org/10.1086/110852. The two-orbit alternative of August 1969.
Peter van de Kamp (1963), Astrometric Study of Barnard's Star from Plates Taken with the 24-inch Sproul Refractor. The Astronomical Journal 68(7), 515-521. doi:10.1086/109001. doi.org/10.1086/109001. The first orbit, for history only.
George Gatewood and Heinrich Eichhorn (1973), An unsuccessful search for a planetary companion of Barnard's star (BD +4°3561). The Astronomical Journal 78(8), 769-776. doi:10.1086/111480. doi.org/10.1086/111480. The deciding control.
John L. Hershey (1973), Astrometric analysis of the field of AC +65°6955 from plates taken with the Sproul 24-inch refractor. The Astronomical Journal 78(5), 421-425. doi:10.1086/111436. doi.org/10.1086/111436. The second layer.
Peter van de Kamp (1975), Astrometric study of Barnard's star from plates taken with the Sproul 61-cm refractor. The Astronomical Journal 80(8), 658-661. doi:10.1086/111791. doi.org/10.1086/111791. The claimant’s reply.
Jieun Choi, Chris McCarthy, Geoffrey W. Marcy, Andrew W. Howard, Debra A. Fischer, John A. Johnson, Howard Isaacson and Jason T. Wright (2013), Precise Doppler Monitoring of Barnard's Star. The Astrophysical Journal 764(2), 131. doi:10.1088/0004-637X/764/2/131. arXiv:1208.2273v2. The Doppler verdict on the old claims.
J. I. González Hernández and 39 coauthors (2024), A sub-Earth-mass planet orbiting Barnard's star. Astronomy & Astrophysics 690, A79. doi:10.1051/0004-6361/202451311. arXiv:2410.00569v1.
Ritvik Basant, Rafael Luque, Jacob L. Bean, Andreas Seifahrt, Madison Brady, Lily L. Zhao, Nina Brown, Tanya Das, Julian Stürmer, David Kasper, Rohan Gupta and Guðmundur Stefánsson (2025), Four Sub-Earth Planets Orbiting Barnard's Star from MAROON-X and ESPRESSO. The Astrophysical Journal Letters 982(1), L1. doi:10.3847/2041-8213/adb8d5. arXiv:2503.08095v1.
Xander Byrne, Claire Marie Guimond, Amy Bonsor, Haiyang S Wang, Sophia R Vaughan and James G Rogers (2026), The Barnard's Star planetary system: stability, composition, and evolution of four sub-Earth exoplanets. Monthly Notices of the Royal Astronomical Society 550(2), stag1207. doi:10.1093/mnras/stag1207. doi.org/10.1093/mnras/stag1207. Its introduction describes Following a controversial history of exoplanet detection claims before treating the four planets as confirmed.
E. E. Mamajek, A. Prsa, G. Torres and 19 others (IAU Inter-Division A-G Working Group on Nominal Units for Stellar and Planetary Astronomy) (2015), IAU 2015 Resolution B3 on Recommended Nominal Conversion Constants for Selected Solar and Planetary Properties. arXiv. arXiv:1510.07674v1. Jupiter is 317.8 Earth masses.

The historical papers are copyrighted by the American Astronomical Society and read through the NASA Astrophysics Data System. This page ships only the numbers it transcribed, the coordinates of two digitised figures, and short quotations, each cited in place; no scan, table layout or figure is reproduced. The licence notes for every file are in LICENCE.txt beside this page.