ABOUT THE SPEAKER
Matthew Carter - Type designer
Even if you don’t recognize his name, chances are you’ve seen Matthew Carter’s work -- his type designs include some of the world’s most familiar digital typefaces.

Why you should listen

MacArthur Fellow Matthew Carter started his career as a punchcutter, a print artisan who physically carves each letter into metal. He had already designed several legendary typefaces (in 1975, he created Bell Centennial for use in phone books) when he stepped into the digital design ring. As part of a project undertaken for Microsoft, Carter created early and successful examples of screen type emphasizing clarity and ease of long-term viewing, including the familiar Verdana, Galliard and Georgia. A recent work is MS Sitka, a family of digital fonts that are built to be readable at many sizes in print and onscreen.

Print and online publications such as Sports Illustrated, Newsweek, Wired and the Washington Post have all commissioned Carter fonts, leading Microsoft's typography blog to call him "the person who shapes the way we read about the world."

Carter is a principal of Carter & Cone Type Inc., a Cambridge, Massachusetts-based company specializing in designing and producing original typefaces. As he says: "A typeface is a beautiful collection of letters, not a collection of beautiful letters."

More profile about the speaker
Matthew Carter | Speaker | TED.com
TED2014

Matthew Carter: My life in typefaces

Filmed:
1,195,279 views

Pick up a book, magazine or screen, and more than likely you'll come across some typography designed by Matthew Carter. In this charming talk, the man behind typefaces such as Verdana, Georgia and Bell Centennial (designed just for phone books -- remember them?), takes us on a spin through a career focused on the very last pixel of each letter of a font.
- Type designer
Even if you don’t recognize his name, chances are you’ve seen Matthew Carter’s work -- his type designs include some of the world’s most familiar digital typefaces. Full bio

Double-click the English transcript below to play the video.

00:12
Type is something we consume
0
709
1526
00:14
in enormous quantities.
1
2235
1922
00:16
In much of the world,
2
4157
1082
00:17
it's completely inescapable.
3
5239
1868
00:19
But few consumers are concerned to know
4
7107
3023
00:22
where a particular typeface came from
5
10130
2332
00:24
or when or who designed it,
6
12462
2764
00:27
if, indeed, there was any human agency involved
7
15226
3353
00:30
in its creation, if it didn't just sort of materialize
8
18579
2797
00:33
out of the software ether.
9
21376
3712
00:37
But I do have to be concerned with those things.
10
25088
3525
00:40
It's my job.
11
28613
1517
00:42
I'm one of the tiny handful of people
12
30130
2210
00:44
who gets badly bent out of shape
13
32340
2114
00:46
by the bad spacing of the T and the E
14
34454
2594
00:49
that you see there.
15
37048
1900
00:50
I've got to take that slide off.
16
38948
1442
00:52
I can't stand it. Nor can Chris.
17
40390
2071
00:54
There. Good.
18
42461
1661
00:56
So my talk is about the connection
19
44122
1768
00:57
between technology and design of type.
20
45890
3400
01:01
The technology has changed
21
49290
2663
01:03
a number of times since I started work:
22
51953
2872
01:06
photo, digital, desktop, screen, web.
23
54825
4661
01:11
I've had to survive those changes and try
24
59486
1738
01:13
to understand their implications for what I do
25
61224
2704
01:15
for design.
26
63928
1391
01:17
This slide is about the effect of tools on form.
27
65319
5170
01:22
The two letters, the two K's,
28
70489
2901
01:25
the one on your left, my right, is modern,
29
73390
3325
01:28
made on a computer.
30
76715
1411
01:30
All straight lines are dead straight.
31
78126
2042
01:32
The curves have that kind of
mathematical smoothness
32
80168
2810
01:34
that the Bézier formula imposes.
33
82978
3746
01:38
On the right, ancient Gothic,
34
86724
2497
01:41
cut in the resistant material of steel by hand.
35
89221
4059
01:45
None of the straight lines are actually straight.
36
93280
2125
01:47
The curves are kind of subtle.
37
95405
2165
01:49
It has that spark of life from the human hand
38
97570
4714
01:54
that the machine or the program
39
102284
1914
01:56
can never capture.
40
104198
1933
01:58
What a contrast.
41
106131
1979
02:00
Well, I tell a lie.
42
108110
2579
02:02
A lie at TED. I'm really sorry.
43
110689
2988
02:05
Both of these were made on a computer,
44
113677
2121
02:07
same software, same Bézier curves,
45
115798
1926
02:09
same font format.
46
117724
1692
02:11
The one on your left
47
119416
2534
02:13
was made by Zuzana Licko at Emigre,
48
121950
2254
02:16
and I did the other one.
49
124204
1435
02:17
The tool is the same, yet the letters are different.
50
125639
3356
02:20
The letters are different
51
128995
1523
02:22
because the designers are different.
52
130518
1206
02:23
That's all. Zuzana wanted hers to look like that.
53
131724
3261
02:26
I wanted mine to look like that. End of story.
54
134985
3121
02:30
Type is very adaptable.
55
138106
2204
02:32
Unlike a fine art, such as sculpture or architecture,
56
140310
3321
02:35
type hides its methods.
57
143631
3399
02:39
I think of myself as an industrial designer.
58
147030
2797
02:41
The thing I design is manufactured,
59
149827
1566
02:43
and it has a function:
60
151393
1876
02:45
to be read, to convey meaning.
61
153269
1838
02:47
But there is a bit more to it than that.
62
155107
1646
02:48
There's the sort of aesthetic element.
63
156753
1848
02:50
What makes these two letters different
64
158601
2716
02:53
from different interpretations by different designers?
65
161317
3032
02:56
What gives the work of some designers
66
164349
2013
02:58
sort of characteristic personal style,
67
166362
2911
03:01
as you might find in the work of a fashion designer,
68
169273
2537
03:03
an automobile designer, whatever?
69
171810
3075
03:06
There have been some cases, I admit,
70
174885
1775
03:08
where I as a designer
71
176660
1150
03:09
did feel the influence of technology.
72
177810
3089
03:12
This is from the mid-'60s,
73
180899
3060
03:15
the change from metal type to photo,
74
183959
2235
03:18
hot to cold.
75
186194
1702
03:19
This brought some benefits
76
187896
1197
03:21
but also one particular drawback:
77
189093
3346
03:24
a spacing system that only provided
78
192439
2673
03:27
18 discrete units for letters
79
195112
3983
03:31
to be accommodated on.
80
199095
2463
03:33
I was asked at this time to design
81
201558
1976
03:35
a series of condensed sans serif types
82
203534
2738
03:38
with as many different variants as possible
83
206272
3163
03:41
within this 18-unit box.
84
209435
3847
03:45
Quickly looking at the arithmetic,
85
213282
1639
03:46
I realized I could only actually make three
86
214921
3309
03:50
of related design. Here you see them.
87
218230
3387
03:53
In Helvetica Compressed, Extra Compressed,
88
221617
2591
03:56
and Ultra Compressed, this rigid 18-unit system
89
224208
3811
04:00
really boxed me in.
90
228019
1573
04:01
It kind of determined the proportions
91
229592
1833
04:03
of the design.
92
231425
2289
04:05
Here are the typefaces, at least the lower cases.
93
233714
4030
04:09
So do you look at these and say,
94
237744
2692
04:12
"Poor Matthew, he had to submit to a problem,
95
240436
3139
04:15
and by God it shows in the results."
96
243575
3807
04:19
I hope not.
97
247382
1307
04:20
If I were doing this same job today,
98
248689
2241
04:22
instead of having 18 spacing units,
99
250930
2826
04:25
I would have 1,000.
100
253756
3084
04:28
Clearly I could make more variants,
101
256840
2453
04:31
but would these three members
of the family be better?
102
259293
4696
04:35
It's hard to say without actually doing it,
103
263989
1852
04:37
but they would not be better in the proportion
104
265841
1707
04:39
of 1,000 to 18, I can tell you that.
105
267548
3089
04:42
My instinct tells you that any improvement
106
270637
1938
04:44
would be rather slight, because they were designed
107
272575
3078
04:47
as functions of the system they were designed to fit,
108
275653
2921
04:50
and as I said, type is very adaptable.
109
278574
2409
04:52
It does hide its methods.
110
280983
2787
04:55
All industrial designers work within constraints.
111
283770
2682
04:58
This is not fine art.
112
286452
2492
05:00
The question is, does a constraint
113
288944
1878
05:02
force a compromise?
114
290822
2667
05:05
By accepting a constraint,
115
293489
1511
05:07
are you working to a lower standard?
116
295000
2449
05:09
I don't believe so, and I've always been encouraged
117
297449
1962
05:11
by something that Charles Eames said.
118
299411
2124
05:13
He said he was conscious of working
119
301535
1545
05:15
within constraints,
120
303080
1038
05:16
but not of making compromises.
121
304118
3226
05:19
The distinction between a constraint
122
307344
2561
05:21
and a compromise is obviously very subtle,
123
309905
2375
05:24
but it's very central to my attitude to work.
124
312280
5671
05:29
Remember this reading experience?
125
317951
2891
05:32
The phone book. I'll hold the slide
126
320842
1425
05:34
so you can enjoy the nostalgia.
127
322267
4535
05:38
This is from the mid-'70s early trials
128
326802
2746
05:41
of Bell Centennial typeface I designed
129
329548
2629
05:44
for the U.S. phone books,
130
332177
1774
05:45
and it was my first experience of digital type,
131
333951
3319
05:49
and quite a baptism.
132
337270
4170
05:53
Designed for the phone books, as I said,
133
341440
1599
05:55
to be printed at tiny size on newsprint
134
343039
3368
05:58
on very high-speed rotary presses
135
346407
2318
06:00
with ink that was kerosene and lampblack.
136
348725
2752
06:03
This is not a hospitable environment
137
351477
3841
06:07
for a typographic designer.
138
355318
3202
06:10
So the challenge for me was to design type
139
358520
1899
06:12
that performed as well as possible
140
360419
1501
06:13
in these very adverse production conditions.
141
361920
4745
06:18
As I say, we were in the infancy of digital type.
142
366665
2855
06:21
I had to draw every character by hand
143
369520
2819
06:24
on quadrille graph paper --
144
372339
1715
06:26
there were four weights of Bell Centennial —
145
374054
1952
06:28
pixel by pixel, then encode
them raster line by raster line
146
376006
3353
06:31
for the keyboard.
147
379359
981
06:32
It took two years, but I learned a lot.
148
380340
4424
06:36
These letters look as though they've been chewed
149
384764
1630
06:38
by the dog or something or other,
150
386394
1444
06:39
but the missing pixels at the intersections
151
387838
1942
06:41
of strokes or in the crotches
152
389780
1605
06:43
are the result of my studying the effects
153
391385
3174
06:46
of ink spread on cheap paper
154
394559
2946
06:49
and reacting, revising the font accordingly.
155
397505
3737
06:53
These strange artifacts are designed to compensate
156
401242
3250
06:56
for the undesirable effects of scale
157
404492
2752
06:59
and production process.
158
407244
2286
07:01
At the outset, AT&T had wanted
159
409530
2846
07:04
to set the phone books in Helvetica,
160
412376
3241
07:07
but as my friend Erik Spiekermann said
161
415617
1665
07:09
in the Helvetica movie, if you've seen that,
162
417282
2503
07:11
the letters in Helvetica were designed to be
163
419785
2035
07:13
as similar to one another as possible.
164
421820
2741
07:16
This is not the recipe for legibility at small size.
165
424561
3274
07:19
It looks very elegant up on a slide.
166
427835
2590
07:22
I had to disambiguate these forms
167
430425
2190
07:24
of the figures as much as possible in Bell Centennial
168
432615
3000
07:27
by sort of opening the shapes up, as you can see
169
435615
2295
07:29
in the bottom part of that slide.
170
437910
2913
07:32
So now we're on to the mid-'80s,
171
440823
2657
07:35
the early days of digital outline fonts,
172
443480
2556
07:38
vector technology.
173
446036
2357
07:40
There was an issue at that time
174
448393
2038
07:42
with the size of the fonts,
175
450431
1856
07:44
the amount of data that was required to find
176
452287
2884
07:47
and store a font in computer memory.
177
455171
4970
07:52
It limited the number of fonts you could get
178
460141
1658
07:53
on your typesetting system at any one time.
179
461799
2990
07:56
I did an analysis of the data,
180
464789
4149
08:00
and found that a typical serif face
181
468938
2524
08:03
you see on the left
182
471462
1459
08:04
needed nearly twice as much data
183
472921
2016
08:06
as a sans serif in the middle
184
474937
2536
08:09
because of all the points required
185
477473
2062
08:11
to define the elegantly curved serif brackets.
186
479535
4508
08:16
The numbers at the bottom of the slide, by the way,
187
484043
3434
08:19
they represent the amount of data
188
487477
1702
08:21
needed to store each of the fonts.
189
489179
3805
08:24
So the sans serif, in the middle,
190
492984
2166
08:27
sans the serifs, was much more economical,
191
495150
2964
08:30
81 to 151.
192
498114
2199
08:32
"Aha," I thought. "The engineers have a problem.
193
500313
3657
08:35
Designer to the rescue."
194
503970
2235
08:38
I made a serif type, you can see it on the right,
195
506205
2347
08:40
without curved serifs.
196
508552
1947
08:42
I made them polygonal, out
of straight line segments,
197
510499
2416
08:44
chamfered brackets.
198
512915
1983
08:46
And look, as economical in data as a sans serif.
199
514898
4368
08:51
We call it Charter, on the right.
200
519266
2299
08:53
So I went to the head of engineering
201
521565
1970
08:55
with my numbers, and I said proudly,
202
523535
2458
08:57
"I have solved your problem."
203
525993
2128
09:00
"Oh," he said. "What problem?"
204
528121
3713
09:03
And I said, "Well, you know, the problem
205
531834
1646
09:05
of the huge data you require
for serif fonts and so on."
206
533480
3417
09:08
"Oh," he said. "We solved that problem last week.
207
536897
3547
09:12
We wrote a compaction routine that reduces
208
540444
2156
09:14
the size of all fonts by an order of magnitude.
209
542600
2580
09:17
You can have as many fonts on your system
210
545180
1988
09:19
as you like."
211
547168
1558
09:20
"Well, thank you for letting me know," I said.
212
548726
2614
09:23
Foiled again.
213
551340
1630
09:24
I was left with a design solution
214
552970
2045
09:27
for a nonexistent technical problem.
215
555015
4473
09:31
But here is where the story sort
of gets interesting for me.
216
559488
2469
09:33
I didn't just throw my design away
217
561957
2595
09:36
in a fit of pique.
218
564552
1401
09:37
I persevered.
219
565953
1737
09:39
What had started as a technical exercise
220
567690
2168
09:41
became an aesthetic exercise, really.
221
569858
3264
09:45
In other words, I had come to like this typeface.
222
573122
3049
09:48
Forget its origins. Screw that.
223
576171
2319
09:50
I liked the design for its own sake.
224
578490
2490
09:52
The simplified forms of Charter
225
580980
2383
09:55
gave it a sort of plain-spoken quality
226
583363
2083
09:57
and unfussy spareness
227
585446
1551
09:58
that sort of pleased me.
228
586997
2490
10:01
You know, at times of technical innovation,
229
589487
2553
10:04
designers want to be influenced
230
592040
1520
10:05
by what's in the air.
231
593560
1736
10:07
We want to respond. We want to be pushed
232
595296
2231
10:09
into exploring something new.
233
597527
3411
10:12
So Charter is a sort of parable for me, really.
234
600938
2899
10:15
In the end, there was no hard and fast causal link
235
603837
3790
10:19
between the technology and the design of Charter.
236
607627
3193
10:22
I had really misunderstood the technology.
237
610820
3762
10:26
The technology did suggest something to me,
238
614582
3328
10:29
but it did not force my hand,
239
617910
2117
10:32
and I think this happens very often.
240
620027
2717
10:34
You know, engineers are very smart,
241
622744
2626
10:37
and despite occasional frustrations
242
625370
1556
10:38
because I'm less smart,
243
626926
1527
10:40
I've always enjoyed working with them
244
628453
1727
10:42
and learning from them.
245
630180
2078
10:44
Apropos, in the mid-'90s,
246
632258
2342
10:46
I started talking to Microsoft
247
634600
2691
10:49
about screen fonts.
248
637291
2388
10:51
Up to that point, all the fonts on screen
249
639679
2421
10:54
had been adapted from previously existing
250
642100
2753
10:56
printing fonts, of course.
251
644853
2377
10:59
But Microsoft foresaw correctly
252
647230
2503
11:01
the movement, the stampede
253
649733
2150
11:03
towards electronic communication,
254
651883
2787
11:06
to reading and writing onscreen
255
654670
2030
11:08
with the printed output as being sort of secondary
256
656700
3133
11:11
in importance.
257
659833
2223
11:14
So the priorities were just tipping at that point.
258
662056
3579
11:17
They wanted a small core set of fonts
259
665635
2194
11:19
that were not adapted but designed for the screen
260
667829
3305
11:23
to face up to the problems of screen,
261
671134
2573
11:25
which were their coarse resolution displays.
262
673707
3842
11:29
I said to Microsoft, a typeface designed
263
677549
3531
11:33
for a particular technology
264
681080
1569
11:34
is a self-obsoleting typeface.
265
682649
3375
11:38
I've designed too many faces in the past
266
686024
2094
11:40
that were intended to mitigate technical problems.
267
688118
3579
11:43
Thanks to the engineers, the
technical problems went away.
268
691697
2835
11:46
So did my typeface.
269
694532
2489
11:49
It was only a stopgap.
270
697021
3131
11:52
Microsoft came back to say that
271
700152
1541
11:53
affordable computer monitors
272
701693
1630
11:55
with better resolutions
273
703323
1197
11:56
were at least a decade away.
274
704520
2596
11:59
So I thought, well, a decade, that's not bad,
275
707116
2654
12:01
that's more than a stopgap.
276
709770
2412
12:04
So I was persuaded, I was convinced,
277
712182
2001
12:06
and we went to work on what became Verdana
278
714183
2322
12:08
and Georgia,
279
716505
1672
12:10
for the first time working not on paper
280
718177
2340
12:12
but directly onto the screen from the pixel up.
281
720517
3960
12:16
At that time, screens were binary.
282
724477
3853
12:20
The pixel was either on or it was off.
283
728330
3040
12:23
Here you see the outline of a letter,
284
731370
2855
12:26
the cap H,
285
734225
1467
12:27
which is the thin black line, the contour,
286
735692
2801
12:30
which is how it is stored in memory,
287
738493
2876
12:33
superimposed on the bitmap,
288
741369
1670
12:35
which is the grey area,
289
743039
2148
12:37
which is how it's displayed on the screen.
290
745187
1837
12:39
The bitmap is rasterized from the outline.
291
747024
3146
12:42
Here in a cap H, which is all straight lines,
292
750170
2241
12:44
the two are in almost perfect sync
293
752411
2088
12:46
on the Cartesian grid.
294
754499
4368
12:50
Not so with an O.
295
758867
3126
12:53
This looks more like bricklaying
than type design,
296
761993
2727
12:56
but believe me, this is a good bitmap O,
297
764720
2901
12:59
for the simple reason that it's symmetrical
298
767621
2015
13:01
in both x and y axes.
299
769636
2480
13:04
In a binary bitmap, you actually can't ask
300
772116
2858
13:06
for more than that.
301
774974
1720
13:08
I would sometimes make, I don't know,
302
776694
2404
13:11
three or four different versions of a difficult letter
303
779098
2356
13:13
like a lowercase A,
304
781454
1506
13:14
and then stand back to choose which was the best.
305
782960
3540
13:18
Well, there was no best,
306
786500
2095
13:20
so the designer's judgment comes in
307
788595
2425
13:23
in trying to decide
308
791020
1389
13:24
which is the least bad.
309
792409
3041
13:27
Is that a compromise?
310
795450
2450
13:29
Not to me, if you are working
311
797900
1546
13:31
at the highest standard the technology will allow,
312
799446
3087
13:34
although that standard may be
313
802533
2209
13:36
well short of the ideal.
314
804742
2479
13:39
You may be able to see on this slide
315
807221
1591
13:40
two different bitmap fonts there.
316
808812
2162
13:42
The "a" in the upper one, I think,
317
810974
1696
13:44
is better than the "a" in the lower one,
318
812670
1913
13:46
but it still ain't great.
319
814583
2576
13:49
You can maybe see the effect better
320
817159
1907
13:51
if it's reduced. Well, maybe not.
321
819066
3376
13:54
So I'm a pragmatist, not an idealist,
322
822442
2396
13:56
out of necessity.
323
824838
1503
13:58
For a certain kind of temperament,
324
826341
1929
14:00
there is a certain kind of satisfaction
325
828270
1700
14:01
in doing something that cannot be perfect
326
829970
3638
14:05
but can still be done to the best of your ability.
327
833608
3869
14:09
Here's the lowercase H from Georgia Italic.
328
837477
4857
14:14
The bitmap looks jagged and rough.
329
842334
2273
14:16
It is jagged and rough.
330
844607
1859
14:18
But I discovered, by experiment,
331
846466
1996
14:20
that there is an optimum slant
332
848462
3217
14:23
for an italic on a screen
333
851679
1946
14:25
so the strokes break well
334
853625
2332
14:27
at the pixel boundaries.
335
855957
2483
14:30
Look in this example how, rough as it is,
336
858440
2781
14:33
how the left and right legs
337
861221
2050
14:35
actually break at the same level.
338
863271
1949
14:37
That's a victory. That's good, right there.
339
865220
3520
14:40
And of course, at the lower depths,
340
868740
3178
14:43
you don't get much choice.
341
871918
1923
14:45
This is an S, in case you were wondering.
342
873841
5045
14:50
Well, it's been 18 years now
343
878886
2154
14:53
since Verdana and Georgia were released.
344
881040
2650
14:55
Microsoft were absolutely right,
345
883690
2090
14:57
it took a good 10 years,
346
885780
2414
15:00
but screen displays now do have
347
888194
2280
15:02
improved spatial resolution,
348
890474
2453
15:04
and very much improved photometric resolution
349
892927
3472
15:08
thanks to anti-aliasing and so on.
350
896399
3154
15:11
So now that their mission is accomplished,
351
899553
3697
15:15
has that meant the demise
352
903250
1730
15:16
of the screen fonts that I designed
353
904980
1880
15:18
for coarser displays back then?
354
906860
2651
15:21
Will they outlive the now-obsolete screens
355
909511
3395
15:24
and the flood of new web fonts
356
912906
2172
15:27
coming on to the market?
357
915078
1503
15:28
Or have they established their own
358
916581
1860
15:30
sort of evolutionary niche
359
918441
2238
15:32
that is independent of technology?
360
920679
3737
15:36
In other words, have they been absorbed
361
924416
1671
15:38
into the typographic mainstream?
362
926087
3313
15:41
I'm not sure, but they've had a good run so far.
363
929400
3658
15:45
Hey, 18 is a good age for anything
364
933058
2495
15:47
with present-day rates of attrition,
365
935553
2151
15:49
so I'm not complaining.
366
937704
1914
15:51
Thank you.
367
939618
2839
15:54
(Applause)
368
942457
2177

▲Back to top

ABOUT THE SPEAKER
Matthew Carter - Type designer
Even if you don’t recognize his name, chances are you’ve seen Matthew Carter’s work -- his type designs include some of the world’s most familiar digital typefaces.

Why you should listen

MacArthur Fellow Matthew Carter started his career as a punchcutter, a print artisan who physically carves each letter into metal. He had already designed several legendary typefaces (in 1975, he created Bell Centennial for use in phone books) when he stepped into the digital design ring. As part of a project undertaken for Microsoft, Carter created early and successful examples of screen type emphasizing clarity and ease of long-term viewing, including the familiar Verdana, Galliard and Georgia. A recent work is MS Sitka, a family of digital fonts that are built to be readable at many sizes in print and onscreen.

Print and online publications such as Sports Illustrated, Newsweek, Wired and the Washington Post have all commissioned Carter fonts, leading Microsoft's typography blog to call him "the person who shapes the way we read about the world."

Carter is a principal of Carter & Cone Type Inc., a Cambridge, Massachusetts-based company specializing in designing and producing original typefaces. As he says: "A typeface is a beautiful collection of letters, not a collection of beautiful letters."

More profile about the speaker
Matthew Carter | Speaker | TED.com