Fact-checked by Grok 2 weeks ago

Haversine formula

The haversine formula is a mathematical used to compute the between two points on the surface of a , such as , given their latitudes and longitudes. It relies on the haversine function, defined as \operatorname{hav}(\theta) = \sin^2(\theta/2) = \frac{1 - \cos \theta}{2}, which is half the of \theta and offers for small angular differences by avoiding in trigonometric computations. This formula is particularly valuable in , , and geographic information systems (GIS) for accurately modeling distances on spherical approximations of planetary bodies, assuming a uniform radius like Earth's mean value of approximately 6,371 km. The haversine function originated in the context of and logarithmic tables for astronomical calculations. It first appeared in printed tables of logarithmic versines compiled by mathematician and astronomer José de Mendoza y Ríos in 1801, facilitating computations for methods in determining at sea. The term "haversine," a contraction of "half-versed-sine," was coined in 1835 by British mathematician and clergyman James Inman in the third edition of his textbook Navigation and Nautical Astronomy, where he introduced it to simplify nautical calculations involving small angles. In its standard form for distance calculation, the haversine formula is expressed as a = \operatorname{hav}(\phi_2 - \phi_1) + \cos \phi_1 \cos \phi_2 \operatorname{hav}(\lambda_2 - \lambda_1), where \phi_1, \phi_2 are the latitudes and \lambda_1, \lambda_2 are the longitudes (in radians), followed by the central angle \Delta \sigma = 2 \operatorname{asin}(\sqrt{a}) and d = R \Delta \sigma, with R as the sphere's radius. This arrangement derives from the but uses haversines to enhance precision for short distances, as highlighted in navigational literature. Though superseded by more precise ellipsoidal models like Vincenty's formula for high-accuracy , the haversine formula remains a foundational tool due to its simplicity and effectiveness for spherical approximations in applications ranging from aviation routing to geolocation.

Fundamentals

Haversine Function

The haversine function, denoted \operatorname{hav}(\theta), is defined mathematically as \operatorname{hav}(\theta) = \sin^2\left(\frac{\theta}{2}\right), where \theta is an angular measure typically expressed in radians. This formulation arises from half-angle identities in and provides a convenient way to express certain spherical trigonometric relations. The haversine is closely related to the function, defined as \operatorname{versin}(\theta) = 1 - \cos(\theta), such that \operatorname{hav}(\theta) = \frac{\operatorname{versin}(\theta)}{2}. This equivalence highlights its connection to cosine-based computations, but the sine-squared form is often preferred for its numerical stability. For \theta \in [0, \pi], the haversine function is continuous, increasing, and bounded between 0 and 1, with \operatorname{hav}(0) = 0 and \operatorname{hav}(\pi) = 1. It proves especially valuable in avoiding errors during small-angle approximations; for instance, when \theta is small, \cos(\theta) \approx 1 - \frac{\theta^2}{2}, but direct subtraction of nearly equal values leads to precision loss, whereas \sin^2(\theta/2) \approx (\theta/2)^2 remains accurate using standard . The haversine function was first utilized in navigational tables by Spanish astronomer José de Mendoza y Ríos in 1801, as documented by mathematician and historian Florian Cajori. The term "haversine," a contraction of "half-versed-sine," was introduced in 1835 by English mathematician James Inman in his textbook Navigation and Nautical Astronomy to aid spherical calculations in astronomy and navigation. This function underpins the haversine formula for great-circle distances between points on a .

Spherical Geometry Basics

Spherical coordinates provide a system for locating points on the surface of a using angular measurements relative to a reference plane and axis. , denoted as φ, measures the angle north or south of the , ranging from -90° at the to +90° at the , while , denoted as λ, measures the angle east or west from the , ranging from -180° to +180°. These coordinates are defined with respect to the center of the sphere, treating and as angles on a of R. The represents the shortest path between two points on a , corresponding to the of a that passes through both points and the sphere's center. This distance is given by d = R \Delta \sigma, where \Delta \sigma is the subtended by the two points at the sphere's center, measured in radians. are the largest possible circles on the sphere, with the same radius as the sphere itself, and they form the basis for and geodesic calculations. In applications involving , the planet is often modeled as a perfect with a mean of approximately 6371 to simplify computations, though this ignores the ellipsoidal shape and variations due to and . This spherical approximation is sufficient for many purposes but introduces minor errors compared to more precise ellipsoidal models. Points on the sphere can be represented as vectors in three-dimensional Cartesian coordinates for facilitating angular separation calculations. For a point at latitude φ and longitude λ on a unit sphere (where R = 1), the position vector is (\cos \phi \cos \lambda, \cos \phi \sin \lambda, \sin \phi). Scaling by the sphere's radius R yields the full coordinates: (R \cos \phi \cos \lambda, R \cos \phi \sin \lambda, R \sin \phi). These vector forms enable dot product computations to determine the central angle \Delta \sigma via \cos \Delta \sigma = \mathbf{p_1} \cdot \mathbf{p_2} / R^2, providing a foundation for distance-related trigonometric functions.

Formulation

Distance Calculation

The Haversine formula computes the , the shortest path along the surface of a between two points specified by their latitudes and longitudes. It is particularly suited for in calculations involving small angular separations, making it valuable in and geospatial applications. The formula assumes a spherical model of the , ignoring its oblateness for simplicity. The required inputs are the latitudes \phi_1 and \phi_2 of the two points, and the absolute difference in longitudes \Delta\lambda = |\lambda_2 - \lambda_1|, with all angles expressed in radians. Longitudes and latitudes are typically provided in degrees and must be converted to radians by multiplying by \pi / 180. The formula proceeds in steps to avoid direct computation of small differences that could lead to loss of precision: \begin{align*} a &= \sin^2\left(\frac{\Delta\phi}{2}\right) + \cos \phi_1 \cdot \cos \phi_2 \cdot \sin^2\left(\frac{\Delta\lambda}{2}\right), \\ c &= 2 \cdot \atan2\left(\sqrt{a}, \sqrt{1 - a}\right), \\ d &= R \cdot c, \end{align*} where R is the radius of the sphere, and the terms \sin^2(\theta/2) correspond to the haversine function \hav(\theta). The output d represents the in the same units as R. For Earth-distance calculations, R is often taken as the mean of 6371 ; alternatively, the equatorial of 6378 may be used for greater accuracy near the . These values derive from precise geodetic measurements.

Law of Haversines

The law of haversines is a fundamental relation in that expresses any side or angle of a spherical triangle in terms of the haversines of the other elements. It applies to spherical triangles formed by the intersections of great circles on a , where the sides represent angular distances and the angles are dihedral angles between the planes of the great circles. A common form of the law for the side c opposite angle C in a spherical triangle with sides a, b and angles A, B is given by \hav c = \hav(a - b) + \sin a \sin b \, \hav C where \hav \theta = \frac{1 - \cos \theta}{2} = \sin^2 \left( \frac{\theta}{2} \right). This form is particularly useful for navigation computations involving known sides and the included angle. Analogous expressions exist for the other sides and for the angles, enabling symmetric treatment of all elements in the triangle. These formulas facilitate solving oblique spherical triangles without directly computing cosines, which is advantageous in tabular computations. Compared to the , the law of haversines offers improved , particularly for small angular distances where cosine values are close to unity, leading to subtractive cancellation errors; the haversine function, being quadratic in small angles, preserves in such cases. This property made it valuable in pre-digital and persists in modern computational for accurate triangle solutions. The haversine distance formula between two points on a arises as a special case of the law of haversines when the spherical triangle degenerates into a configuration defined by the points' latitudes and difference. This extends the basic two-point distance calculation to more complex spherical geometries involving multiple great-circle arcs.

Derivation

Geometric Proof

To derive the Haversine formula geometrically, represent the positions of two points on a as unit vectors in three-dimensional Cartesian coordinates, assuming a spherical model of . Let the points P_1 and P_2 have latitudes \phi_1, \phi_2 and longitudes \lambda_1, \lambda_2, where latitudes range from -\pi/2 to \pi/2 and longitudes from -\pi to \pi. The corresponding unit vectors are given by: \begin{align*} \mathbf{P_1} &= \begin{pmatrix} \cos\phi_1 \cos\lambda_1 \\ \cos\phi_1 \sin\lambda_1 \\ \sin\phi_1 \end{pmatrix}, \\ \mathbf{P_2} &= \begin{pmatrix} \cos\phi_2 \cos\lambda_2 \\ \cos\phi_2 \sin\lambda_2 \\ \sin\phi_2 \end{pmatrix}. \end{align*} The central angle \Delta\sigma subtended by the arc between P_1 and P_2 at the sphere's center equals the angle between the vectors \mathbf{P_1} and \mathbf{P_2}, which is found using the dot product formula for unit vectors: \cos \Delta\sigma = \mathbf{P_1} \cdot \mathbf{P_2}. Expanding the dot product explicitly produces the spherical law of cosines in terms of latitude and longitude difference \Delta\lambda = \lambda_2 - \lambda_1: \cos \Delta\sigma = \sin \phi_1 \sin \phi_2 + \cos \phi_1 \cos \phi_2 \cos \Delta\lambda. To arrive at the Haversine formula, apply the half-angle identity for the haversine function, defined as \hav \theta = \sin^2(\theta/2) = (1 - \cos \theta)/2: $1 - \cos \Delta\sigma = 2 \sin^2 \left( \frac{\Delta\sigma}{2} \right), so \hav(\Delta\sigma) = \frac{1 - \cos \Delta\sigma}{2}. Substituting the expanded cosine expression and applying trigonometric half-angle identities \sin^2(\alpha/2) = (1 - \cos \alpha)/2 to the latitude difference \Delta\phi = \phi_2 - \phi_1 and longitude difference terms yields: \hav(\Delta\sigma) = \sin^2 \left( \frac{\Delta\phi}{2} \right) + \cos \phi_1 \cos \phi_2 \sin^2 \left( \frac{\Delta\lambda}{2} \right). This equation gives the haversine of the directly in terms of the input coordinates, avoiding issues with cosine near unity for small angles. Geometrically, this derivation corresponds to the spherical formed by the two points P_1, P_2, and the N, where the sides from N to P_1 and N to P_2 are the co-latitudes \pi/2 - \phi_1 and \pi/2 - \phi_2, the angle at N is \Delta\lambda, and the side P_1P_2 has length \Delta\sigma. The approach embeds this in the plane through the sphere's center, using the to capture the angular separation.

Trigonometric Derivation

The trigonometric derivation of the haversine formula relies on the spherical law of cosines, a fundamental relation in spherical trigonometry for any spherical triangle with sides a, b, c (in angular measure) and opposite angles A, B, C: \cos c = \cos a \cos b + \sin a \sin b \cos C This equation, established in classical treatments of spherical geometry, allows computation of one side given the other two sides and the included angle. To derive the great-circle distance between two points on a sphere specified by latitudes \phi_1 and \phi_2 (in radians) and longitude difference \Delta\lambda, construct a spherical triangle with vertices at the two points and the north pole. The sides adjacent to the north pole are the co-latitudes: a = \frac{\pi}{2} - \phi_1 and b = \frac{\pi}{2} - \phi_2. The angle C at the north pole equals \Delta\lambda. Substituting these into the spherical law of cosines yields: \cos c = \sin\phi_1 \sin\phi_2 + \cos\phi_1 \cos\phi_2 \cos\Delta\lambda Here, c represents the central angle subtended by the great-circle arc between the points; multiplying by the sphere's radius R gives the distance d = R c. This form directly follows from the trigonometric substitutions \cos a = \sin\phi_1, \sin a = \cos\phi_1, and analogously for b. To express this in haversine form, apply the haversine identity \hav \theta = \frac{1 - \cos \theta}{2} = \sin^2 \frac{\theta}{2}, which rearranges the law of cosines into a form suited for navigational tables and avoiding subtraction of near-equal quantities for small angles. The general identity transforming the spherical law of cosines to haversines is the law of haversines: \hav c = \hav(a - b) + \sin a \sin b \hav C Substituting a = \frac{\pi}{2} - \phi_1, b = \frac{\pi}{2} - \phi_2, and C = \Delta\lambda simplifies to the haversine formula, since \sin a = \cos \phi_1, \sin b = \cos \phi_2, \hav(a - b) = \hav(\phi_2 - \phi_1), and \hav C = \hav(\Delta\lambda): \hav c = \hav(\phi_2 - \phi_1) + \cos\phi_1 \cos\phi_2 \hav(\Delta\lambda) This equation, a special case of the broader law of haversines in spherical trigonometry, facilitates accurate distance computation, particularly for small separations where cosine-based forms suffer numerical instability.

Applications and Examples

The Haversine formula has been historically employed in to compute great-circle distances between observed celestial bodies and a vessel's position, facilitating sight reductions that integrate with techniques for position fixes prior to the advent of GPS. In , navigators used the formula to estimate distances traveled based on course, speed, and time, often in conjunction with haversine tables for logarithmic computations to minimize errors over long voyages. Its adoption extended to route planning in the early , where it enabled pilots to calculate efficient great-circle paths for transoceanic flights, reducing fuel consumption and flight time on models. In modern applications, the Haversine formula serves as a foundational method for distance calculations between positions obtained from GPS systems, such as in smartphones, by determining great-circle distances between latitude-longitude coordinates. Many mapping applications use the Haversine formula for straight-line distance calculations and proximity queries, often in conjunction with APIs like for displaying results on approximations. Within geodesy, the formula approximates distances on a spherical model of Earth, treating the planet as a perfect sphere with a mean radius of approximately 6371 km, which simplifies computations for global-scale analyses but introduces errors when compared to more precise ellipsoidal representations. For high-precision requirements, such as those using the WGS84 ellipsoid standard in satellite geodesy, the Haversine formula's limitations necessitate alternatives like Vincenty's formula, which accounts for Earth's oblate shape and yields geodesic distances accurate to within millimeters over long baselines. Error analysis reveals that the Haversine formula achieves an accuracy of about 0.3% to 0.5% relative to true distances on , making it suitable for most navigational purposes but less ideal for sub-millimeter geodetic surveys. Its numerical stability is particularly advantageous for small distances, where it avoids subtraction cancellation errors inherent in other spherical trigonometric approaches, ensuring reliable results even for separations under 1 .

Computational Implementation

The computational implementation of the Haversine formula involves converting coordinates to radians, computing the haversine of the , and applying the inverse haversine to obtain the . A standard outline for calculating the distance d between two points with latitudes \phi_1, \phi_2 and longitudes \lambda_1, \lambda_2 (in degrees) on a of R is as follows:
function haversine_distance(φ1_deg, λ1_deg, φ2_deg, λ2_deg, [R](/page/R)):
    # Convert degrees to radians
    φ1 = φ1_deg * π / 180
    λ1 = λ1_deg * π / 180
    φ2 = φ2_deg * π / 180
    λ2 = λ2_deg * π / 180
    
    # Differences in radians
    Δφ = φ2 - φ1
    Δλ = λ2 - λ1
    
    # Haversine computation
    a = sin²(Δφ / 2) + cos(φ1) * cos(φ2) * sin²(Δλ / 2)
    
    # Inverse haversine using [atan2](/page/Atan2) for [angular distance](/page/Angular_distance) c
    c = 2 * [atan2](/page/Atan2)(√a, √(1 - a))
    
    # [Distance](/page/Distance)
    d = R * c
    return d
This implementation relies on the two-argument arctangent function (atan2) to ensure the correct quadrant and avoid division by zero in the inverse haversine step. The Haversine formula offers improved numerical stability compared to the spherical law of cosines, particularly for points in close proximity where the central angle is small. In the law of cosines, computing \cos c = \cos a \cos b + \sin a \sin b \cos C involves subtracting two nearly equal large values (close to 1) when distances are short, leading to catastrophic cancellation and loss of precision in floating-point arithmetic. By contrast, the Haversine approach uses half-angle sine terms, which remain well-conditioned even as \Delta\phi and \Delta\lambda approach zero, preserving accuracy down to sub-millimeter scales on Earth's surface. In programming practice, implementations should employ double-precision (64-bit) to minimize rounding errors, as single-precision (32-bit) may introduce inaccuracies exceeding 1 meter for global distances. Special handling is required for antipodal points, where the \Delta\sigma = \pi; here, \sin(\Delta\lambda / 2) = 0 and \sin(\Delta\phi / 2) = 0, causing the argument to atan2 to be (0, 0) and yielding NaN—in such cases, the distance can be set directly to \pi R. Additionally, input validation for ([-90^\circ, 90^\circ]) and ([-180^\circ, 180^\circ]) ranges is recommended to prevent extraneous results. The Haversine formula is supported in various programming libraries and geospatial tools. In , the haversine package provides a straightforward for computation using the . For , the geolib library includes Haversine-based calculations suitable for web applications.

Numerical Example

To illustrate the application of the Haversine , consider calculating the between at 40.7128° N, 74.0060° W and at 51.5074° N, 0.1278° W. The for the c between two points with latitudes \phi_1, \phi_2 and longitudes \lambda_1, \lambda_2 (in radians) is given by a = \sin^2\left(\frac{\Delta\phi}{2}\right) + \cos\phi_1 \cos\phi_2 \sin^2\left(\frac{\Delta\lambda}{2}\right), c = 2 \arcsin(\sqrt{a}), where \Delta\phi = \phi_2 - \phi_1 and \Delta\lambda = \lambda_2 - \lambda_1, followed by the d = R c using Earth's mean radius R = 6371 km. Convert the coordinates to radians: \phi_1 \approx 0.7111, \phi_2 \approx 0.8988, \lambda_1 \approx -1.2922, \lambda_2 \approx -0.0022. Then, \Delta\phi \approx 0.1877 and \Delta\lambda \approx 1.2900. Compute \sin^2(\Delta\phi/2) \approx 0.0088, \cos\phi_1 \approx 0.7580, \cos\phi_2 \approx 0.6216, and \sin^2(\Delta\lambda/2) \approx 0.3625. Thus, a \approx 0.0088 + (0.7580 \times 0.6216 \times 0.3625) \approx 0.0088 + 0.1709 = 0.1797. Next, \sqrt{a} \approx 0.4238, so c = 2 \arcsin(0.4238) \approx 2 \times 0.437 = 0.874 radians. The distance is d \approx 6371 \times 0.874 \approx 5570 km. This result aligns closely with the established great-circle distance of approximately 5569 km between these points, with the minor discrepancy attributable to the formula's assumption of a perfect sphere rather than Earth's oblate spheroid shape. For greater precision in specific regions, R can be adjusted: use the equatorial radius of 6378 km for paths near the equator or the polar radius of 6357 km for polar routes.

References

  1. [1]
    [PDF] Distance between Points on the Earth's Surface - KSU Math
    which, in principle, concludes this narrative. The distance is often represented in terms of the so-called haversine function, defined by haversin A = sin.Missing: explanation | Show results with:explanation
  2. [2]
    Haversine -- from Wolfram MathWorld
    The haversine is a trigonometric function defined by hav(z) = 1/2vers(z) = 1/2(1-cosz) = sin^2(1/2z), where versin(z) is the versine.
  3. [3]
    Computing Distances - NYU Computer Science
    Haversine Formula (from R.W. Sinnott, "Virtues of the Haversine", Sky and Telescope, vol. 68, no. 2, 1984, p. 159): dlon = lon2 - lon1 dlat = lat2 - lat1 a ...
  4. [4]
    Dictionary of National Biography, 1885-1900/Inman, James
    Dec 28, 2020 · In the third edition (1835) he introduced a new trigonometrical function, the half-versine, or haversine, the logarithms of which were added ...
  5. [5]
    Latitude and Longitude - PWG Home - NASA
    Oct 10, 2016 · 8a. The Horizon. Any location on Earth can be identified by two numbers--its latitude and its longitude. If a pilot or a ship's captain wants ...
  6. [6]
    Lab 1 - University of Hawaii System
    The global coordinate system is known as latitude and longitude. These are spherical coordinates, based on angles rather than distances, which can be used ...
  7. [7]
    Great Circle -- from Wolfram MathWorld
    A great circle is a section of a sphere that contains a diameter of the sphere (Kern and Bland 1948, p. 87). Sections of the sphere that do not contain a ...
  8. [8]
    Spherical Trigonometry and Navigation - Stony Brook University
    If the sphere has radius R, so will all its great circles; so the length of an arc is exactly R times the radian measure of the central angle it intersects (or ...
  9. [9]
    Earth Fact Sheet - the NSSDCA - NASA
    Missing: authoritative | Show results with:authoritative
  10. [10]
    [PDF] Chapter 2. Vectors and the geographic coordinate system
    Positive longitude is measured east of the. Prime Meridian; positive latitude is measured north of the Equator. The Cartesian coordinates of a point Q, whose ...
  11. [11]
    Calculate distance and bearing between two Latitude/Longitude ...
    Haversine formula: · d = R ⋅ c ; where: φ is latitude, λ is longitude, R is earth's radius (mean radius = 6,371km); note that angles need to be in radians to pass ...Missing: authoritative | Show results with:authoritative
  12. [12]
    Planetary Physical Parameters - JPL Solar System Dynamics
    Table Column Descriptions. Equatorial Radius, Radius of the planet at the equator. Mean Radius, Radius of a sphere with the equivalent volume of the planet.Missing: authoritative source
  13. [13]
    Spherical Trigonometry -- from Wolfram MathWorld
    471), and the haversine formula for angles is given by. havA, = (sin(s-b)sin(s-c))/(sinbsinc). (46). = (hava-hav(b-c))/(sinbsinc). (47). = hav[pi-(B+C)]+ ...
  14. [14]
    [PDF] From Navigation to Star Hopping: Forgotten Formulae
    Astronomy for Sea-men, in 1835, Sir James Inman in- troduced half-verse-sine as haversine, basically for calcu- lating distances on the sea, simplifying the ...
  15. [15]
    [PDF] Geometry and Trigonometry - Professor Terje Haukaas
    Apr 7, 2024 · Using the notation from above the law of haversines states the spherical ... hav(c) hav(θ) = sin2 θ. 2. ⎛. ⎝⎜. ⎞. ⎠⎟. = 1−cos(θ). 2 cos(a ...
  16. [16]
    [PDF] Implementing efficient geo queries for the SPARQL engine QLever
    Mar 3, 2025 · cosines or the law of haversines. But for small angles, the haversine function provides numerical stability, while the cosine function does not.
  17. [17]
    [PDF] The great circle distance - Underground Mathematics
    We now offer four approaches to deducing the haversine formula. The first two of these use coordinates, so we begin with this. Page 2 of 7. Copyright © ...
  18. [18]
    The Cosine-Haversine Formula - jstor
    (7) cos a = cos b cos c + sin b sin c cos A. Since tables of natural haversines and their logarithms are available, the cosine formula may be rearranged in more ...Missing: source | Show results with:source
  19. [19]
    [PDF] An analysis of the development of celestial navigation
    Using his dead reckoning latitude, which he realized was unre- liable, he ... The cosine-haversine formula hav z = hav (L-^dj-f- cos L cos d hav t.<|separator|>
  20. [20]
    [PDF] Instrumentation Milestones In Aerial Celestial Navigation
    This formula may be applied to the. Navigational Triangle to solve for the unknown side. To simplify their calculations, navigators have employed several ...
  21. [21]
    The Hidden Math Behind Air Travel - by fliteX - Aviation Redux
    Mar 16, 2023 · The haversine formula is commonly used to calculate the great-circle distance between two points on a sphere, such as the Earth. It takes into ...
  22. [22]
    Vincenty solutions of geodesics on the ellipsoid - Movable-type.co.uk
    Calculations based on a spherical earth model, such as the (much simpler) Haversine, are accurate to around 0.3% – which is still good enough for many (most?) ...
  23. [23]
    Fast geodesic approximations with Cheap Ruler | by Mapbox
    it's an oblate ellipsoid. Consequently, the Haversine formula can result in an error of up to 0.5%.
  24. [24]
    Distance on a sphere: The Haversine Formula - Esri Community
    Oct 5, 2017 · For example, haversine(θ) = sin²(θ/2). The haversine formula is a very accurate way of computing distances between two points on the surface of ...
  25. [25]
    haversine - PyPI
    Haversine. Calculate the distance (in various units) between two points on Earth using their latitude and longitude.Missing: postgis | Show results with:postgis
  26. [26]
    New York City, NY, USA - Latitude and Longitude Finder
    The latitude of New York City, NY, USA is 40.730610, and the longitude is -73.935242. New York City, NY, USA is located at United States country in the Cities ...
  27. [27]
    Where is London, UK on Map Lat Long Coordinates
    London, UK is located at United Kingdom country in the Cities place category with the gps coordinates of 51° 30' 35.5140'' N and 0° 7' 5.1312'' W.
  28. [28]
    earth_oblate_spheroid.html - UNLV Physics
    Earth Equatorial Radius R_eq_⊕ = 6378.1370 km. ; Earth polar radius R_po_⊕= 6356.7523 km. ; Earth mean radius R_me_⊕ = 6371.0088 km by the Earth conventional mean ...