Figures play a significant role in the expression of scientific ideas. A single well-prepared figure can contribute immeasurably to the clarity of the text, and high standards of presentation are therefore particularly important…

(3) Graphs should be self-explanatory, their purpose evident without reference to the text. Indicate clearly what is being plotted, in both the horizontal and the vertical directions. Include appropriate units. Orient letters and numbers so that they may be easily read from the bottom or the righthand side of the graph. Relevant nongraphic material, such as the key to the symbolism in the graph, may be included within the confines of the graph frame if it will fit without crowding; otherwise put the explanatory material in the caption.

In captions, use available symbols (see Appendix F) to represent data points, but use words to identify curves (for example, “solid,” “dashed,” “dotted,” “dot-dashed,” etc.). A better alternative is to label curves with letters (A, B, etc. ) and to refer to them by letter in the caption (“Curve A represents…” ).

The notation used in graphs should be standard and consistent with the notation used in the text. Write 0.1, not .1, 0 1, or 0,1. Do not capitalize letters indiscriminately: write

in units of q/a, not IN UNITS OF q/a

and

kinetic energy Ef(meV) not KINETIC ENERGY Ef (MEV)

Take care to preserve standard forms for symbols and abbreviations, as you would in text. Standard units should be well spaced off and enclosed in parentheses.

If possible, do not use powers of ten in axis labels: use instead the appropriate prefixes of the Systeme International (see Table IV). If powers of ten must be used, write for example

R( 10-4[Omega]) or 104R ([Omega]).

Never write

R x 10-4[Omega] or R /10-4[Omega],

because in these forms it is not clear whether the scale numbers have been or are to be multiplied by 10-4. Better still, attach the power of ten to the largest number on the axis: 8 x 10-4.

Whenever possible, use integer numbers on the axis scales of figures (1, 2, 3, or O, 5, 10, not 1.58, 3.16, 4.75 or 1.5, 3.0, 4.5). If this is not feasible, then there must be a number both before and after the decimal point: Use 0.5, not .5, and 5, not 5., etc. Do not use unnecessary decimal places: 1.0, 1.5, 2.0 is acceptable, but not 1.00, 2.00, 3.00.

Coordinate ruling should be limited in number to those necessary to guide the eye in making a reading to the desired degree of approximation. Ticks to indicate coordinate values, placed within all four sides of the graph, increase readability, and are recommended. Closely spaced coordinate rulings are appropriate only for computation charts. It is often impossible in a journal to make a graph large enough to preserve accuracy of the data beyond two significant figures. If that accuracy is not sufficient for your purposes, present the data as a table.

Graphs with large blank areas, or large areas containing only nongraphic material, are unacceptable; use only the ranges of coordinates for which there are data. If similar quantities are plotted several times, use shifted ordinate scales for each plot and enclose the plots in one large rectangle, not in separate boxes, thereby saving space. Isometric drawings giving the illusion of three dimensions to the family of curves are often better.

(4) In diagrams of electrical circuits, the values of resistances, inductances, etc. and component designations should be lettered directly on the diagram. A separate parts list in the caption is then unnecessary, except for special or unusual components.

(6) Computer-drawn figures can now be made equal in quality to those drawn by a skilled draftsmen, and the same criteria should apply to them. In particular, lines should be dark, and of adequate width to survive reduction. Lettering should be simple, pleasing to the eye, in one typeface only and no more than two sizes. The slash through a zero to differentiate it from the letter O is unacceptable.

Joining every pair of adjacent experimental points is an easy solution but it may lead to curves that are too obviously a series of line segments or that are very “noisy.” It is preferable to produce a curve by some smoother method, such as by the use of an analytical approximation, in which the calculated points may be as close as desired and only the lines joining them need appear.