1
00:00:00,000 --> 00:00:01,940
Welcome back to the Deep Dive. This is where

2
00:00:01,940 --> 00:00:04,440
we take a whole stack of sources and, well, give

3
00:00:04,440 --> 00:00:06,799
you the shortcut to understanding the core ideas.

4
00:00:07,040 --> 00:00:09,019
That's right. Today, we're diving into something

5
00:00:09,019 --> 00:00:11,660
pretty complex, something fundamental to how

6
00:00:11,660 --> 00:00:14,939
many languages work. Tone. We're looking at the

7
00:00:14,939 --> 00:00:18,179
acoustic and cognitive architecture behind it.

8
00:00:18,239 --> 00:00:20,320
Yeah, and it's fascinating because if you speak

9
00:00:20,320 --> 00:00:23,030
a language like English or maybe Spanish... You

10
00:00:23,030 --> 00:00:25,750
definitely use pitch. We all do. But we use it

11
00:00:25,750 --> 00:00:29,269
for intonation, showing emotion, asking a question

12
00:00:29,269 --> 00:00:31,789
versus making a statement. Right. That feels

13
00:00:31,789 --> 00:00:34,329
universal. It pretty much is. But that's fundamentally

14
00:00:34,329 --> 00:00:37,740
different from how... what, maybe 70 % of the

15
00:00:37,740 --> 00:00:40,539
world's languages use pitch, which is tone, using

16
00:00:40,539 --> 00:00:43,200
pitch changes to actually distinguish words themselves.

17
00:00:43,479 --> 00:00:45,700
And that's exactly our mission for you, the listener,

18
00:00:45,840 --> 00:00:48,100
today. We've got a deep dive into the source

19
00:00:48,100 --> 00:00:50,140
material lined up. We'll start right down at

20
00:00:50,140 --> 00:00:51,960
the physics of sound production and work our

21
00:00:51,960 --> 00:00:54,950
way up to some pretty high -level theories. The

22
00:00:54,950 --> 00:00:57,310
goal really is to give you a fast but thorough

23
00:00:57,310 --> 00:00:59,969
understanding of what tone actually is, how it

24
00:00:59,969 --> 00:01:01,929
functions in language, and maybe what it tells

25
00:01:01,929 --> 00:01:04,709
us about, you know, the complexity of human language

26
00:01:04,709 --> 00:01:07,430
itself. It sounds like a big undertaking. Moving

27
00:01:07,430 --> 00:01:10,790
from just sound waves to actual linguistic meaning.

28
00:01:10,989 --> 00:01:13,709
It is. And that's why we absolutely have to start

29
00:01:13,709 --> 00:01:16,930
by clarifying three key terms that honestly get

30
00:01:16,930 --> 00:01:18,590
mixed up all the time. We need to build that

31
00:01:18,590 --> 00:01:20,769
bridge from the acoustic signal to the linguistic

32
00:01:20,769 --> 00:01:23,620
meaning. Step by step. Okay, let's start at the

33
00:01:23,620 --> 00:01:27,260
absolute base level. The pure physics. The sound

34
00:01:27,260 --> 00:01:29,299
itself. Right. That's fundamental frequency.

35
00:01:29,700 --> 00:01:31,700
We're usually written as five dollars. It's purely

36
00:01:31,700 --> 00:01:34,500
acoustic. You measure it in hertz. Hertz. Okay.

37
00:01:34,599 --> 00:01:37,120
And it's simply the rate your vocal folds vibrate.

38
00:01:37,159 --> 00:01:39,120
How many times they cycle open and closed per

39
00:01:39,120 --> 00:01:41,620
second. Faster vibration means higher few dollars

40
00:01:41,620 --> 00:01:43,719
neck. That's it. It's a physical measurement.

41
00:01:43,780 --> 00:01:45,840
Totally independent of whether anyone's listening

42
00:01:45,840 --> 00:01:47,719
or what language they speak. Then we bring in

43
00:01:47,719 --> 00:01:51,060
the human element. The listener. Exactly. That

44
00:01:51,060 --> 00:01:54,180
gets us to pitch. Pitch is perceptual. It's what

45
00:01:54,180 --> 00:01:56,659
you, the hearer, actually hear. That subjective

46
00:01:56,659 --> 00:01:59,680
feeling of a sound being high or low. So it tracks

47
00:01:59,680 --> 00:02:02,260
via the dollars, but it's not identical. Precisely.

48
00:02:02,260 --> 00:02:04,920
Pitch generally follows 7 .Rs, but it's not a

49
00:02:04,920 --> 00:02:07,319
perfect one -to -one mapping. Our perception

50
00:02:07,319 --> 00:02:09,800
of pitch can be influenced by other things, like

51
00:02:09,800 --> 00:02:12,759
volume amplitude, or even the sounds that came

52
00:02:12,759 --> 00:02:15,719
just before it. It's subjective experience. Okay,

53
00:02:15,819 --> 00:02:18,960
physical fault of dollars, perceived pitch. And

54
00:02:18,960 --> 00:02:21,180
now the crucial one for today, the linguistic

55
00:02:21,180 --> 00:02:24,159
one. And that's tone. Tone is the properly linguistic

56
00:02:24,159 --> 00:02:26,919
concept. It's defined by its job in the language,

57
00:02:27,139 --> 00:02:30,439
its functional load. Does a difference in pitch

58
00:02:30,439 --> 00:02:33,460
change the meaning of a word or a grammatical

59
00:02:33,460 --> 00:02:36,400
form? Like your Cantonese example. Exactly. In

60
00:02:36,400 --> 00:02:38,860
Cantonese, saying yo with a high -level pitch,

61
00:02:38,979 --> 00:02:41,979
let's say yo -h means worry. But say the exact

62
00:02:41,979 --> 00:02:44,759
same syllable, yo, with a low -level pitch, yo,

63
00:02:44,919 --> 00:02:47,539
and it means again. Wow. Same sounds, different

64
00:02:47,539 --> 00:02:49,520
words, just based on pitch. Completely different

65
00:02:49,520 --> 00:02:52,879
words. The segments Y, A, U are identical. It's

66
00:02:52,879 --> 00:02:55,919
the tone, the contrast of use of pitch that distinguishes

67
00:02:55,919 --> 00:02:58,500
them as separate items in the lexicon. And that

68
00:02:58,500 --> 00:03:00,039
functional difference is what we're really digging

69
00:03:00,039 --> 00:03:03,599
into today. Got it. So tone is about meaning,

70
00:03:03,840 --> 00:03:06,900
not just emotion or questions. Once we have that

71
00:03:06,900 --> 00:03:09,300
clear, let's look under the hood. How do we actually

72
00:03:09,300 --> 00:03:12,560
produce these precise pitch differences? What's

73
00:03:12,560 --> 00:03:15,210
the physical machinery? Okay, yeah. It all happens

74
00:03:15,210 --> 00:03:17,889
in the larynx, the voice box. The key structures

75
00:03:17,889 --> 00:03:21,289
there are cartilages, the cricoid at the base,

76
00:03:21,469 --> 00:03:24,430
the thyroid, that's the Adam's apple, and then

77
00:03:24,430 --> 00:03:27,430
the two little arytenoid cartilages at the back.

78
00:03:27,509 --> 00:03:29,349
And the vocal folds stretch between some of those.

79
00:03:29,509 --> 00:03:31,169
They stretch between the thyroid cartilage at

80
00:03:31,169 --> 00:03:33,270
the front and the arytenoid cartilages at the

81
00:03:33,270 --> 00:03:35,729
back. So changing the pitch means physically

82
00:03:35,729 --> 00:03:38,009
changing the state of those vocal folds. Right.

83
00:03:38,090 --> 00:03:41,409
How do you do that? Tension. Mass. Both, essentially.

84
00:03:41,960 --> 00:03:44,699
For pitch raising, the main player is the cricothyroid

85
00:03:44,699 --> 00:03:47,400
muscle. When it contracts, it pulls the thyroid

86
00:03:47,400 --> 00:03:49,439
cartilage forward and down, sort of tilting it.

87
00:03:49,520 --> 00:03:51,460
Okay. And that stretches the vocal folds, makes

88
00:03:51,460 --> 00:03:53,580
them longer and thinner. This stretching does

89
00:03:53,580 --> 00:03:56,280
two crucial things. It decreases their effective

90
00:03:56,280 --> 00:03:59,259
mass and, more importantly, drastically increases

91
00:03:59,259 --> 00:04:01,759
their stiffness or tension. And skiffer things

92
00:04:01,759 --> 00:04:04,800
vibrate faster. Exactly. Faster vibration, higher

93
00:04:04,800 --> 00:04:08,319
$5, higher perceived pitch. The cricothyroid

94
00:04:08,319 --> 00:04:10,699
is like the pitch accelerator. Studies clearly

95
00:04:10,699 --> 00:04:13,560
show its activity ramps up just before and during

96
00:04:13,560 --> 00:04:15,659
a pitch peak. Okay, so that's the accelerator.

97
00:04:15,939 --> 00:04:18,300
What about the brakes? How do you lower the pitch?

98
00:04:18,420 --> 00:04:21,120
Is it just relaxing that muscle? Pitch lowering

99
00:04:21,120 --> 00:04:23,740
is actually a bit more complicated. Yes, it involves

100
00:04:23,740 --> 00:04:26,279
relaxing the cricothyroid, reducing that tension,

101
00:04:26,480 --> 00:04:28,980
but that's often not enough. And just relaxing

102
00:04:28,980 --> 00:04:31,620
might make the voice unstable or breathy. So

103
00:04:31,620 --> 00:04:35,180
another muscle usually gets involved. The thyroarytenoid

104
00:04:35,180 --> 00:04:37,420
muscle. When it contracts, it actually bunches

105
00:04:37,420 --> 00:04:39,680
up the vocal folds, making them shorter and thicker.

106
00:04:39,819 --> 00:04:42,079
Increasing their mouth. Precisely. More mass

107
00:04:42,079 --> 00:04:44,819
means slower vibration, lower viadollars. So

108
00:04:44,819 --> 00:04:47,459
it's this coordinated dance. Relax the stretcher

109
00:04:47,459 --> 00:04:49,459
muscle, contract the thickener muscle. That gives

110
00:04:49,459 --> 00:04:51,879
you controlled low pitch. This anatomical setup

111
00:04:51,879 --> 00:04:54,120
gives us the physical range, that continuous

112
00:04:54,120 --> 00:04:57,019
gradient viadollar curve we can produce. But

113
00:04:57,019 --> 00:04:59,540
now we hit that really tough question. You mentioned

114
00:04:59,540 --> 00:05:02,360
the core of phonology, maybe. How does that smooth,

115
00:05:02,420 --> 00:05:05,259
continuous physical signal get interpreted by

116
00:05:05,259 --> 00:05:07,459
the brain as discrete, meaningful categories

117
00:05:07,459 --> 00:05:10,339
in a language? Yes. This is the phonetic -phonological

118
00:05:10,339 --> 00:05:13,500
divide, and it's absolutely central. The actual

119
00:05:13,500 --> 00:05:16,720
acoustic output, the $50 curve you could measure

120
00:05:16,720 --> 00:05:19,199
with the machine, is always gradient. It slides

121
00:05:19,199 --> 00:05:21,620
up and down across the speaker's range. Right.

122
00:05:21,680 --> 00:05:24,139
But the mental representation, the phonology

123
00:05:24,139 --> 00:05:26,449
of the language... operates with categories,

124
00:05:26,769 --> 00:05:29,750
discrete units like high tone H or low tone L.

125
00:05:29,870 --> 00:05:32,569
It's like converting an analog signal to a digital

126
00:05:32,569 --> 00:05:35,810
one. So how do we as linguists or even as listeners

127
00:05:35,810 --> 00:05:38,050
tell the difference? When does a pitch change

128
00:05:38,050 --> 00:05:41,339
just? you know, a physical wobble? And when is

129
00:05:41,339 --> 00:05:43,920
it a signal that the language intends? Why should

130
00:05:43,920 --> 00:05:45,600
the listener care about this difference? You

131
00:05:45,600 --> 00:05:48,240
should care because your brain has to care. One

132
00:05:48,240 --> 00:05:50,980
type of variation is essentially noise predictable

133
00:05:50,980 --> 00:05:53,319
physical effects you need to filter out. The

134
00:05:53,319 --> 00:05:55,660
other type is the meaningful signal you need

135
00:05:55,660 --> 00:05:57,680
to decode. Okay, explain that. What makes something

136
00:05:57,680 --> 00:05:59,680
phonetic noise? Phonetic effects are typically

137
00:05:59,680 --> 00:06:02,220
variable and context dependent. They depend on

138
00:06:02,220 --> 00:06:04,600
things like how fast you're talking, the inertia

139
00:06:04,600 --> 00:06:06,720
of your vocal folds. It takes a tiny fraction

140
00:06:06,720 --> 00:06:08,620
of a second to actually hit the target pitch

141
00:06:08,620 --> 00:06:11,639
and co -articulation how neighboring sounds influence

142
00:06:11,639 --> 00:06:14,300
the pitch. Like it's harder to hit a really high

143
00:06:14,300 --> 00:06:16,699
tone right after a really low one. Exactly. There's

144
00:06:16,699 --> 00:06:19,420
a lag, a transition. And crucially, listeners

145
00:06:19,420 --> 00:06:22,959
are incredibly good at unconsciously compensating

146
00:06:22,959 --> 00:06:26,600
for these predictable phonetic variations. We

147
00:06:26,600 --> 00:06:29,800
sort of hear through the noise. to get to the

148
00:06:29,800 --> 00:06:32,620
intended underlying category. Whereas a phonological

149
00:06:32,620 --> 00:06:35,540
effect is different. It's not noise. Right. Phonological

150
00:06:35,540 --> 00:06:37,759
effects are categorical. Think of them like flipping

151
00:06:37,759 --> 00:06:40,379
a switch. It's either on or off, high or low.

152
00:06:40,839 --> 00:06:43,560
If a rule applies consistently, regardless of

153
00:06:43,560 --> 00:06:45,839
speech rate, and results in a change that could

154
00:06:45,839 --> 00:06:48,540
potentially distinguish meaning, even if it doesn't

155
00:06:48,540 --> 00:06:50,980
in that specific word, it's likely phonological.

156
00:06:51,220 --> 00:06:54,360
The phonology provides the target category, and

157
00:06:54,360 --> 00:06:56,699
the phonetics just executes it as best it can

158
00:06:56,699 --> 00:06:58,810
within the physical constraints. This seems like

159
00:06:58,810 --> 00:07:01,069
a good place to bring in those three terms that

160
00:07:01,069 --> 00:07:04,649
describe pitch falling across an utterance. Declination,

161
00:07:04,870 --> 00:07:08,029
downdrift, and downstep. They show this gradient

162
00:07:08,029 --> 00:07:10,829
to categorical path, right? Perfectly. They illustrate

163
00:07:10,829 --> 00:07:13,509
the client from universal physics to language

164
00:07:13,509 --> 00:07:15,730
-specific grammar. Start with the universal one.

165
00:07:15,930 --> 00:07:19,370
That's declination. It's a general gradual phonetic

166
00:07:19,370 --> 00:07:22,410
lowering of the overall pitch baseline across

167
00:07:22,410 --> 00:07:25,870
an utterance. It happens in pretty much all languages,

168
00:07:26,110 --> 00:07:29,110
tonal or not. Why does it happen? It's mostly

169
00:07:29,110 --> 00:07:32,670
tied to breath control. As you speak, your subglottal

170
00:07:32,670 --> 00:07:35,129
air pressure tends to decrease slightly, which

171
00:07:35,129 --> 00:07:37,389
naturally lowers the baseline pitch. It's just

172
00:07:37,389 --> 00:07:39,089
a physical consequence of running out of steam,

173
00:07:39,250 --> 00:07:41,709
essentially. Okay, so that's just physics. When

174
00:07:41,709 --> 00:07:44,350
does it become linguistic? It becomes linguistic

175
00:07:44,350 --> 00:07:47,189
when the language hijacks that tendency and makes

176
00:07:47,189 --> 00:07:49,149
it part of the grammar. That's downdrift. So

177
00:07:49,149 --> 00:07:51,649
it's phonologized declination. Exactly. Downdrift

178
00:07:51,649 --> 00:07:54,920
isn't universal. It's a specific rule in certain

179
00:07:54,920 --> 00:07:56,959
languages, and it's triggered by a particular

180
00:07:56,959 --> 00:08:00,040
phonological element, almost always a low L tone.

181
00:08:00,259 --> 00:08:03,000
How does it work? In a language with downdrift,

182
00:08:03,040 --> 00:08:05,019
when you have a sequence like high -low -high,

183
00:08:05,180 --> 00:08:07,939
H -L -H, that second H tone is automatically

184
00:08:07,939 --> 00:08:10,240
realized phonetically lower than the first H

185
00:08:10,240 --> 00:08:12,879
tone. The L tone actively pulls the pitch register

186
00:08:12,879 --> 00:08:15,100
down for everything that follows it within a

187
00:08:15,100 --> 00:08:17,439
certain domain. Okay, so the L tone forces a

188
00:08:17,439 --> 00:08:20,120
step down. What about downstep, then? That sounds

189
00:08:20,120 --> 00:08:23,120
similar. It is similar, but... Even more abstract.

190
00:08:23,360 --> 00:08:25,500
This is where it gets tricky, especially in many

191
00:08:25,500 --> 00:08:28,560
African languages. Downstep, sometimes called

192
00:08:28,560 --> 00:08:31,139
non -automatic downstep, is when you get that

193
00:08:31,139 --> 00:08:33,980
same lowering effect on a high tone. But without

194
00:08:33,980 --> 00:08:36,700
an immediately preceding low tone. Exactly. The

195
00:08:36,700 --> 00:08:40,419
register steps down, but there's no overt L tone

196
00:08:40,419 --> 00:08:43,639
on the surface to trigger it. The cause is analyzed

197
00:08:43,639 --> 00:08:46,299
as a floating low tone. A floating tone, like

198
00:08:46,299 --> 00:08:48,990
a tone that's not attached to any vowel. Precisely.

199
00:08:49,029 --> 00:08:51,929
It's phonologically there in the underlying representation.

200
00:08:52,070 --> 00:08:54,230
It's part of the word structure, but it doesn't

201
00:08:54,230 --> 00:08:57,830
get associated with any specific TBU tone -bearing

202
00:08:57,830 --> 00:09:01,289
unit, usually a vowel or mora. It's phonologically

203
00:09:01,289 --> 00:09:04,190
active but phonetically invisible itself. Okay,

204
00:09:04,250 --> 00:09:06,409
hold on. This is where I, and maybe the listener,

205
00:09:06,490 --> 00:09:08,629
might push back a bit. If the low tone isn't

206
00:09:08,629 --> 00:09:10,509
actually pronounced, how do we know it's there?

207
00:09:10,649 --> 00:09:12,669
Why not just say this language has a weird rule?

208
00:09:13,049 --> 00:09:15,870
Lower the pitch after this specific morpheme

209
00:09:15,870 --> 00:09:18,950
or something. That's a fair challenge. The reason

210
00:09:18,950 --> 00:09:22,090
we posit the floating L is twofold. First, the

211
00:09:22,090 --> 00:09:24,389
amount of lowering it causes is often identical

212
00:09:24,389 --> 00:09:26,970
to the lowering caused by a fully realized L

213
00:09:26,970 --> 00:09:29,610
tone in the same language, downdrift. It behaves

214
00:09:29,610 --> 00:09:32,870
like an L tone. Okay. Second, downstep often

215
00:09:32,870 --> 00:09:35,309
occurs precisely where we know, historically

216
00:09:35,309 --> 00:09:37,870
or through other alternations, that a vowel carrying

217
00:09:37,870 --> 00:09:41,120
an L tone used to be but got deleted. The segment

218
00:09:41,120 --> 00:09:43,539
disappeared, but its tone was left floating behind,

219
00:09:43,820 --> 00:09:46,779
still able to influence its neighbors. Ah, so

220
00:09:46,779 --> 00:09:49,200
the floating tone explains why the lowering happens

221
00:09:49,200 --> 00:09:51,860
there and not somewhere else. It provides the

222
00:09:51,860 --> 00:09:54,659
structural motivation. Exactly. Without the floating

223
00:09:54,659 --> 00:09:57,399
L, the downstep looks arbitrary, like magic.

224
00:09:57,620 --> 00:09:59,919
With it, we see it as a regular phonological

225
00:09:59,919 --> 00:10:02,200
process to instability after segment deletion.

226
00:10:02,320 --> 00:10:04,879
This floating L can then cause iterative lowering.

227
00:10:05,120 --> 00:10:07,720
Each floating L encountered steps the whole pitch

228
00:10:07,720 --> 00:10:09,860
register down again. That makes sense. It's about

229
00:10:09,860 --> 00:10:11,860
finding the underlying system rather than just

230
00:10:11,860 --> 00:10:15,090
describing surface patterns. Okay, let's shift

231
00:10:15,090 --> 00:10:16,909
gears. We've got the mechanics and the basic

232
00:10:16,909 --> 00:10:19,250
phonetic phonology split. Now let's look at the

233
00:10:19,250 --> 00:10:22,409
big picture. Where is tone found, and how complex

234
00:10:22,409 --> 00:10:24,730
can these systems get? Right, the global distribution.

235
00:10:24,909 --> 00:10:28,509
As we said, maybe 70 % of languages. The hotspots

236
00:10:28,509 --> 00:10:30,730
are definitely sub -Saharan Africa, particularly

237
00:10:30,730 --> 00:10:34,169
lighter Congo languages. Also, large parts of

238
00:10:34,169 --> 00:10:36,250
the Americas, especially the Otomangian family

239
00:10:36,250 --> 00:10:38,830
in Mexico. And of course, East and Southeast

240
00:10:38,830 --> 00:10:42,470
Asia think Sino -Tibetan, Mon -Khmer, Thai -Kadai

241
00:10:42,470 --> 00:10:44,779
languages. And we mentioned Europe is mostly

242
00:10:44,779 --> 00:10:47,919
non -tonal, with those few exceptions like Swedish

243
00:10:47,919 --> 00:10:50,340
or Serbo -Croatian that are sometimes called

244
00:10:50,340 --> 00:10:53,080
pitch accent languages instead. Yeah, the key

245
00:10:53,080 --> 00:10:54,980
difference there is often inventory size and

246
00:10:54,980 --> 00:10:58,019
function. True tone languages usually have tones

247
00:10:58,019 --> 00:11:00,399
operating contrastively on almost every syllable

248
00:11:00,399 --> 00:11:03,419
or word. Pitch accent systems tend to have maybe

249
00:11:03,419 --> 00:11:06,019
one or two contrastive pitch patterns per word,

250
00:11:06,120 --> 00:11:08,159
often linked specifically to the stress syllable.

251
00:11:08,379 --> 00:11:11,100
It's a bit less pervasive. Let's talk inventory

252
00:11:11,100 --> 00:11:14,000
size then. How many distinct tones can a language

253
00:11:14,000 --> 00:11:17,039
actually juggle? Well, the absolute maximum that's

254
00:11:17,039 --> 00:11:19,559
been reliably reported as phonologically contrastive

255
00:11:19,559 --> 00:11:22,159
is five distinct pitch levels. Languages like

256
00:11:22,159 --> 00:11:24,340
Dan, spoken in West Africa, are cited for this.

257
00:11:24,480 --> 00:11:26,840
But even four contrastive levels are pretty rare.

258
00:11:27,059 --> 00:11:29,360
So simpler systems are more common. Much more

259
00:11:29,360 --> 00:11:32,220
common. The most widespread type is the two -tone

260
00:11:32,220 --> 00:11:37,059
system, just high versus low, HVSL. Vast numbers

261
00:11:37,059 --> 00:11:39,340
of African languages work like this. But you

262
00:11:39,340 --> 00:11:40,840
mentioned something interesting earlier, that

263
00:11:40,840 --> 00:11:43,840
a simple H -L system can actually lead to three

264
00:11:43,840 --> 00:11:45,480
different pitches being heard on the surface.

265
00:11:45,740 --> 00:11:48,419
Ah, yes, that's a key point about the phonetics

266
00:11:48,419 --> 00:11:50,419
-phonology interaction again. Take a language

267
00:11:50,419 --> 00:11:53,860
like Yoruba. The underlying phonological inventory

268
00:11:53,860 --> 00:11:57,820
seems to just be H and L. That's it. Okay. But

269
00:11:57,820 --> 00:12:00,059
the language also allows some syllables to be

270
00:12:00,059 --> 00:12:02,700
underlyingly toneless. They don't have an H or

271
00:12:02,700 --> 00:12:05,500
an L specified in the lexicon. When you actually

272
00:12:05,500 --> 00:12:08,019
speak, though, every syllable needs some pitch.

273
00:12:08,639 --> 00:12:10,799
So these toneless syllables often surface with

274
00:12:10,799 --> 00:12:13,639
a mid -pitch, M. Where does the mid -pitch come

275
00:12:13,639 --> 00:12:16,159
from? It's usually derived phonetically, often

276
00:12:16,159 --> 00:12:18,559
by interpolation between surrounding tones, or

277
00:12:18,559 --> 00:12:20,940
sometimes it's just a default neutral pitch realization

278
00:12:20,940 --> 00:12:24,220
for unspecified syllables. So you have two contrastive

279
00:12:24,220 --> 00:12:27,580
phonological tones, HL, but three distinct phonetic

280
00:12:27,580 --> 00:12:29,940
pitch levels heard, high, mid, low. Very cool.

281
00:12:30,039 --> 00:12:32,399
What about languages that actually contrast three

282
00:12:32,399 --> 00:12:35,649
levels phonologically? Three -tone systems, HML,

283
00:12:35,909 --> 00:12:37,590
are also quite common, found across different

284
00:12:37,590 --> 00:12:40,590
language families globally. Noop in Nigeria is

285
00:12:40,590 --> 00:12:43,450
one example. Punjabi, an Indo -Aryan language,

286
00:12:43,610 --> 00:12:45,990
is another good case. Punjabi? How does that

287
00:12:45,990 --> 00:12:48,529
work? On stressed syllables, Punjabi clearly

288
00:12:48,529 --> 00:12:51,570
contrasts three levels. You get words like kara

289
00:12:51,570 --> 00:12:54,909
with a low tone, meaning horse, kara with a mid

290
00:12:54,909 --> 00:12:57,870
-tone, meaning whip, and kara with a high tone,

291
00:12:58,009 --> 00:13:01,009
meaning leper. The segments are identical, only

292
00:13:01,009 --> 00:13:03,460
the pitch level distinguishes them. So managing

293
00:13:03,460 --> 00:13:05,539
three levels is definitely possible. Four is

294
00:13:05,539 --> 00:13:08,460
rare. Five is the max. Yeah. Got it. Now, how

295
00:13:08,460 --> 00:13:10,299
do linguists actually write these tones down?

296
00:13:10,399 --> 00:13:12,200
You mentioned different systems. Yeah. There

297
00:13:12,200 --> 00:13:14,500
are two main traditions, largely split by geography.

298
00:13:14,820 --> 00:13:17,159
The accent system is dominant in descriptions

299
00:13:17,159 --> 00:13:20,179
of African and American languages. It uses diacritic

300
00:13:20,179 --> 00:13:22,639
marks directly on the vowel. Like accents in

301
00:13:22,639 --> 00:13:25,299
French or Spanish, but for tone? Sort of, but

302
00:13:25,299 --> 00:13:27,100
with dedicated meanings. Typically, an acute

303
00:13:27,100 --> 00:13:30,500
accent means high tone, like ah. A grave accent

304
00:13:30,500 --> 00:13:34,450
means... Low tone, A. A macron often means mid

305
00:13:34,450 --> 00:13:36,950
-tone. What about tones that change pitch on

306
00:13:36,950 --> 00:13:39,529
one syllable, like a fall or rise? Those are

307
00:13:39,529 --> 00:13:42,230
contour tones, and the accent system uses combinations.

308
00:13:42,870 --> 00:13:45,370
A circumflex might be used for a falling tone,

309
00:13:45,610 --> 00:13:49,350
HL, like R, and a caron, or HH, for a rising

310
00:13:49,350 --> 00:13:53,309
tone, LH, like A. It's pretty intuitive for systems

311
00:13:53,309 --> 00:13:55,250
with just a few level tones and maybe simple

312
00:13:55,250 --> 00:13:58,190
contours. But it gets messy if you have lots

313
00:13:58,190 --> 00:14:00,549
of levels or complex contours. It can, yeah.

314
00:14:01,100 --> 00:14:02,840
Which is partly why the other system developed,

315
00:14:02,980 --> 00:14:05,259
mainly for Asian languages, especially Chinese

316
00:14:05,259 --> 00:14:07,720
dialects with their complex tonal systems. That's

317
00:14:07,720 --> 00:14:10,100
the numerical system. Based on numbers. Yes,

318
00:14:10,139 --> 00:14:12,940
usually associated with the linguist Y .R. Chow,

319
00:14:13,059 --> 00:14:16,019
sometimes called Chow tone letters. It imagines

320
00:14:16,019 --> 00:14:18,320
the speaker's pitch range divided into five levels,

321
00:14:18,440 --> 00:14:20,559
where one is the lowest and five is the highest.

322
00:14:20,899 --> 00:14:23,500
A tone is then described by one or more numbers

323
00:14:23,500 --> 00:14:26,039
indicating its pitch path. A high -level tone

324
00:14:26,039 --> 00:14:29,879
would be 55. A low -level tone, 11. A high -falling

325
00:14:29,879 --> 00:14:33,440
tone might be 51. A dipping tone, like the Mandarin

326
00:14:33,440 --> 00:14:36,620
third tone, could be 214. That seems very precise.

327
00:14:36,740 --> 00:14:39,100
You can capture quite complex shapes. It offers

328
00:14:39,100 --> 00:14:41,460
great phonetic detail, absolutely. But it has

329
00:14:41,460 --> 00:14:43,840
a significant drawback, especially for comparing

330
00:14:43,840 --> 00:14:46,000
across speakers or dialects. What's that? The

331
00:14:46,000 --> 00:14:48,279
numbers 1 to 5 are relative to that specific

332
00:14:48,279 --> 00:14:51,500
speaker's pitch range at that moment. My 5 -a

333
00:14:51,500 --> 00:14:53,419
-lo might be acoustically lower than her 4 -fun.

334
00:14:53,879 --> 00:14:57,629
So a tone transcribed as 24, rising. in one dialect

335
00:14:57,629 --> 00:15:00,490
might actually occupy the same relative pitch

336
00:15:00,490 --> 00:15:03,389
space as a tone transcribed as 13 also rising

337
00:15:03,389 --> 00:15:06,149
in another dialect spoken by someone with a different

338
00:15:06,149 --> 00:15:09,029
overall range ah so the numbers look different

339
00:15:09,029 --> 00:15:10,889
but the tones might be functionally the same

340
00:15:10,889 --> 00:15:14,269
relative rise Exactly. The numerical system captures

341
00:15:14,269 --> 00:15:17,450
phonetic shape well, but it can obscure the underlying

342
00:15:17,450 --> 00:15:20,570
phonological categories, especially things like

343
00:15:20,570 --> 00:15:23,090
whether a tone belongs to a higher or lower register.

344
00:15:23,330 --> 00:15:25,809
That's a key theoretical distinction the numbers

345
00:15:25,809 --> 00:15:28,269
alone don't always reveal reliably. Okay, that

346
00:15:28,269 --> 00:15:30,009
makes sense. Let's tackle those contour tones

347
00:15:30,009 --> 00:15:32,330
more directly now. The rises, falls, dips within

348
00:15:32,330 --> 00:15:34,610
a single syllable. The big question seems to

349
00:15:34,610 --> 00:15:38,009
be, are they single, complex units, or are they

350
00:15:38,009 --> 00:15:40,090
just sequences of simpler level tones packed

351
00:15:40,090 --> 00:15:42,139
together? That's a major theoretical debate.

352
00:15:42,320 --> 00:15:45,039
First, some observations. Falling tones, like

353
00:15:45,039 --> 00:15:47,980
HL51, are way more common across the world's

354
00:15:47,980 --> 00:15:50,720
languages than rising tones, LHU15. Interesting.

355
00:15:50,860 --> 00:15:52,799
Why might that be? We'll get to that. It relates

356
00:15:52,799 --> 00:15:56,279
to production difficulty. Also, complex contours,

357
00:15:56,279 --> 00:15:58,600
ones that change direction twice, like a fall

358
00:15:58,600 --> 00:16:03,799
-rise, HLH515, or a rise -fall, LHL151, are extremely

359
00:16:03,799 --> 00:16:07,159
rare, especially in African systems. So the dominant

360
00:16:07,159 --> 00:16:10,309
idea is the sequence hypothesis. The contours

361
00:16:10,309 --> 00:16:12,330
are just level tones stringed together. That's

362
00:16:12,330 --> 00:16:14,929
generally the starting assumption, yes. A falling

363
00:16:14,929 --> 00:16:17,909
tone is underlyingly H followed by L. A rising

364
00:16:17,909 --> 00:16:20,190
tone is L followed by H. What's the strongest

365
00:16:20,190 --> 00:16:22,370
evidence for breaking them down like that? It

366
00:16:22,370 --> 00:16:25,009
often comes down to syllable structure, specifically

367
00:16:25,009 --> 00:16:28,309
the moraic constraint. The idea is that each

368
00:16:28,309 --> 00:16:31,850
underlying tone, H or L, needs something to dock

369
00:16:31,850 --> 00:16:35,899
onto, a tone -bearing unit, TBU. In many languages,

370
00:16:36,059 --> 00:16:38,440
the TBU isn't the whole syllable, but the mora.

371
00:16:38,559 --> 00:16:40,879
Remind us what a mora is. It's a unit of syllable

372
00:16:40,879 --> 00:16:43,639
weight. Roughly, a short vowel has one mora.

373
00:16:43,679 --> 00:16:46,019
A long vowel, or a vowel followed by certain

374
00:16:46,019 --> 00:16:48,220
consonants, like nasals or liquids sometimes,

375
00:16:48,519 --> 00:16:51,320
has two moras. A syllable with two moras is called

376
00:16:51,320 --> 00:16:54,139
hefi. Okay, so how does that relate to contours?

377
00:16:54,509 --> 00:16:56,850
If a contour tone is really a sequence of two

378
00:16:56,850 --> 00:16:59,610
level tones, like H plus L, then according to

379
00:16:59,610 --> 00:17:01,490
this hypothesis, it should generally require

380
00:17:01,490 --> 00:17:04,710
a two mora, heavy syllable, to host both tone

381
00:17:04,710 --> 00:17:07,990
units. A short, single mora, light syllable should

382
00:17:07,990 --> 00:17:10,849
only be able to host one tone unit, meaning only

383
00:17:10,849 --> 00:17:13,210
level tones. And do we see languages behaving

384
00:17:13,210 --> 00:17:15,269
like that? We do. There are many examples. A

385
00:17:15,269 --> 00:17:17,289
really compelling one comes from Wang Sian Chinese.

386
00:17:18,170 --> 00:17:20,789
Syllables ending in a stop consonant, like P

387
00:17:20,789 --> 00:17:23,900
-T -K, are considered short, monomaraic. And

388
00:17:23,900 --> 00:17:26,619
guess what? They can only bear level tones. But

389
00:17:26,619 --> 00:17:29,599
there's a specific process, a type of reduplication,

390
00:17:29,660 --> 00:17:32,019
where one of these stop -final syllables needs

391
00:17:32,019 --> 00:17:35,240
to bear a contour tone phonologically. The language

392
00:17:35,240 --> 00:17:37,920
solves this dilemma in a fascinating way. The

393
00:17:37,920 --> 00:17:40,619
final stop consonant changes into a nasal consonant,

394
00:17:40,839 --> 00:17:44,039
like M -N -OR. Nasals can be moraic, so changing

395
00:17:44,039 --> 00:17:46,599
the consonant makes the syllable heady, bimoraic,

396
00:17:46,619 --> 00:17:48,519
and now it can successfully carry the required

397
00:17:48,519 --> 00:17:51,690
contour tone. Wow. The sound change is purely

398
00:17:51,690 --> 00:17:54,150
driven by the need to make room for the second

399
00:17:54,150 --> 00:17:57,210
part of the contour tone. Exactly. It's powerful

400
00:17:57,210 --> 00:17:59,369
evidence that the contour is being treated as

401
00:17:59,369 --> 00:18:01,750
a sequence of two things needing two docking

402
00:18:01,750 --> 00:18:04,509
spots. Okay. That's strong support for the sequence

403
00:18:04,509 --> 00:18:06,690
idea. But you mentioned there's still a unit

404
00:18:06,690 --> 00:18:09,049
hypothesis. Some people argue contours are single

405
00:18:09,049 --> 00:18:11,569
units sometimes. Yes. The debate isn't completely

406
00:18:11,569 --> 00:18:13,970
settled. In some languages, particularly certain

407
00:18:13,970 --> 00:18:17,509
Asian languages with very stable contour contrasts,

408
00:18:17,509 --> 00:18:19,829
there isn't always clear evidence for decomposition.

409
00:18:20,410 --> 00:18:22,930
And sometimes contours behave phonologically

410
00:18:22,930 --> 00:18:26,650
like single, albeit complex, entities. What kind

411
00:18:26,650 --> 00:18:29,130
of behavior suggests they're units? One argument

412
00:18:29,130 --> 00:18:31,950
comes back to phonetic difficulty. Producing

413
00:18:31,950 --> 00:18:35,910
a rising tone, L plus H, is articulatorily harder

414
00:18:35,910 --> 00:18:38,250
and takes slightly longer than producing a falling

415
00:18:38,250 --> 00:18:41,430
tone, H plus L. Why is that? It involves more

416
00:18:41,430 --> 00:18:43,849
complex muscle coordination to initiate the rise

417
00:18:43,849 --> 00:18:46,529
from a low point. Some languages seem to restrict

418
00:18:46,529 --> 00:18:49,289
rising tones, banning them even on heavy syllables

419
00:18:49,289 --> 00:18:51,349
where a falling tone would be perfectly fine.

420
00:18:51,730 --> 00:18:54,410
If the rise was just L followed by H, you might

421
00:18:54,410 --> 00:18:57,289
expect it anywhere a sequence is allowed. Treating

422
00:18:57,289 --> 00:19:00,150
it as a single, marked, complex unit helps explain

423
00:19:00,150 --> 00:19:02,730
why it might face special restrictions. So the

424
00:19:02,730 --> 00:19:05,630
jury might still be out. Or maybe both analyses

425
00:19:05,630 --> 00:19:07,730
are needed, depending on the language. That's

426
00:19:07,730 --> 00:19:10,190
often the case in linguistics. The sequence hypothesis

427
00:19:10,190 --> 00:19:12,670
is usually the default, but we have to be open

428
00:19:12,670 --> 00:19:15,190
to evidence that points towards unit status in

429
00:19:15,190 --> 00:19:17,730
specific cases. Okay, this complexity clearly

430
00:19:17,730 --> 00:19:20,069
requires some sophisticated theoretical tools.

431
00:19:20,710 --> 00:19:22,670
Let's move into Section 3 and talk about the

432
00:19:22,670 --> 00:19:26,250
architecture of tonal theory itself. How do linguists

433
00:19:26,250 --> 00:19:29,710
model these systems formally, starting with tonal

434
00:19:29,710 --> 00:19:32,730
features? If languages max out maybe four or

435
00:19:32,730 --> 00:19:35,950
five levels, how can we represent those contrasts

436
00:19:35,950 --> 00:19:39,150
efficiently using binary features like phonologists

437
00:19:39,150 --> 00:19:41,589
love to do? Right. The goal is always explanatory

438
00:19:41,589 --> 00:19:44,789
economy. Since a four -level contrast seems to

439
00:19:44,789 --> 00:19:47,230
be the practical upper limit for robust systems,

440
00:19:47,470 --> 00:19:50,289
a system with just two binary features is generally

441
00:19:50,289 --> 00:19:52,769
sufficient. What are the features? The most common

442
00:19:52,769 --> 00:19:55,710
proposal uses register and pitch. Register is

443
00:19:55,710 --> 00:19:57,789
usually seen as the more dominant feature, dividing

444
00:19:57,789 --> 00:19:59,970
the overall pitch space into two main bands,

445
00:20:00,130 --> 00:20:03,569
plus upper register and S upper register. Okay,

446
00:20:03,670 --> 00:20:06,490
a high band and a low band. Exactly. And then

447
00:20:06,490 --> 00:20:08,710
within each register band, the second feature,

448
00:20:08,950 --> 00:20:11,789
pitch plus high, distinguishes two levels. So

449
00:20:11,789 --> 00:20:14,170
that gives us four possible combinations. Precisely.

450
00:20:14,170 --> 00:20:17,390
You get one plus upper plus high, the highest

451
00:20:17,390 --> 00:20:20,690
tone, let's call it H or H plus. two plus upper

452
00:20:20,690 --> 00:20:23,809
high a lower tone within the high band often

453
00:20:23,809 --> 00:20:27,750
called mid or m plus i or lowered high three

454
00:20:27,750 --> 00:20:30,990
upper plus high a higher tone within the low

455
00:20:30,990 --> 00:20:33,450
band often called raised low or m or sometimes

456
00:20:33,450 --> 00:20:37,569
also mid four upper high The lowest tone, L or

457
00:20:37,569 --> 00:20:39,549
L, that neatly generates up to four distinct

458
00:20:39,549 --> 00:20:41,930
levels. It also seems to capture why mid -tones

459
00:20:41,930 --> 00:20:44,009
sometimes pattern with highs and sometimes with

460
00:20:44,009 --> 00:20:46,869
lows. It can, yes. The two high tones, plus upper

461
00:20:46,869 --> 00:20:49,349
high and upper plus high, would naturally fall

462
00:20:49,349 --> 00:20:51,109
somewhere in the middle of the total pitch range,

463
00:20:51,250 --> 00:20:53,009
potentially behaving differently depending on

464
00:20:53,009 --> 00:20:55,730
whether the primary register feature aligns them

465
00:20:55,730 --> 00:20:58,230
with the upper or lower set. Now, the source

466
00:20:58,230 --> 00:21:00,730
material mentioned Duanmu's idea connecting these

467
00:21:00,730 --> 00:21:03,009
abstract features back to the actual larynx muscles

468
00:21:03,009 --> 00:21:04,950
we talked about earlier. Yes, that's a really

469
00:21:04,950 --> 00:21:08,039
appealing aspect of this feature system. Duanmu

470
00:21:08,039 --> 00:21:11,599
proposed a potential link. Register plus upper

471
00:21:11,599 --> 00:21:14,079
might correlate with the tension of the vocalist

472
00:21:14,079 --> 00:21:17,000
muscle, part of the thyroarytenoid. Plus upper

473
00:21:17,000 --> 00:21:20,400
could mean stiff vocal folds, potentially leading

474
00:21:20,400 --> 00:21:23,740
to a clear or even creaky voice quality, while

475
00:21:23,740 --> 00:21:26,819
upper could mean slack folds, perhaps leading

476
00:21:26,819 --> 00:21:29,630
to a breathy voice. And pitch. The pitch feature

477
00:21:29,630 --> 00:21:32,630
plus high would then correlate more directly

478
00:21:32,630 --> 00:21:34,730
with the activity of the cricothyroid muscle,

479
00:21:34,869 --> 00:21:37,049
the primary stretcher, essentially controlling

480
00:21:37,049 --> 00:21:40,150
how thin or thick the folds are vibrating. What's

481
00:21:40,150 --> 00:21:42,210
the advantage of linking features to specific

482
00:21:42,210 --> 00:21:45,730
muscle groups like that? Beyond grounding the

483
00:21:45,730 --> 00:21:48,029
abstract features in phonetics, it offers an

484
00:21:48,029 --> 00:21:50,650
elegant explanation for something quite counterintuitive.

485
00:21:51,640 --> 00:21:53,619
Overlapping contours. Overlapping contours. What

486
00:21:53,619 --> 00:21:55,539
does that mean? It means you could have, say,

487
00:21:55,619 --> 00:21:57,759
a falling tone that starts high but ends quite

488
00:21:57,759 --> 00:22:01,039
low, maybe represented phonetically as 52. And

489
00:22:01,039 --> 00:22:02,759
in the same language, you could have a rising

490
00:22:02,759 --> 00:22:05,779
tone that starts low but ends mid -range, maybe

491
00:22:05,779 --> 00:22:08,720
24. Notice that the endpoint of the high register

492
00:22:08,720 --> 00:22:12,420
fall, 2, is actually lower on the 1 to 5 scale

493
00:22:12,420 --> 00:22:14,859
than the endpoint of the low register rise, 4.

494
00:22:15,420 --> 00:22:17,579
If pitch were just one linear scale, this would

495
00:22:17,579 --> 00:22:19,819
be weird. Right, the ranges seem to cross over.

496
00:22:19,980 --> 00:22:22,500
But if register and pitch are controlled by potentially

497
00:22:22,500 --> 00:22:25,619
independent muscle systems, it makes sense. The

498
00:22:25,619 --> 00:22:28,240
52 fall happens entirely within the plus upper

499
00:22:28,240 --> 00:22:31,140
register. Maybe plus upper plus high going to

500
00:22:31,140 --> 00:22:34,299
plus upper high. The 24 rise happens entirely

501
00:22:34,299 --> 00:22:37,119
within the upper register. Maybe upper high going

502
00:22:37,119 --> 00:22:39,180
to upper plus high. They operate in different

503
00:22:39,180 --> 00:22:41,599
compartments of the voice, allowing their phonetic

504
00:22:41,599 --> 00:22:43,910
ranges to overlap. That's a really neat way to

505
00:22:43,910 --> 00:22:46,430
think about it. Okay, so we have features. Now,

506
00:22:46,430 --> 00:22:49,210
how do we structure them? Why did phonology need

507
00:22:49,210 --> 00:22:52,450
a big revolution, like autosegmental phonology,

508
00:22:52,549 --> 00:22:55,170
specifically for tone? What was wrong with the

509
00:22:55,170 --> 00:22:58,289
old way? The old way, standard generative phonology

510
00:22:58,289 --> 00:23:01,009
in the 60s and 70s treated all features, including

511
00:23:01,009 --> 00:23:03,789
tone, as if they were just properties inherent

512
00:23:03,789 --> 00:23:06,430
to individual segments, like consonants and vowels.

513
00:23:07,109 --> 00:23:09,369
A high tone feature was part of the feature bundle

514
00:23:09,369 --> 00:23:12,150
for a vowel, just like plus back or plus round.

515
00:23:12,309 --> 00:23:15,710
And it didn't work for tone. It failed spectacularly

516
00:23:15,710 --> 00:23:18,690
for tone because tone does things that segmental

517
00:23:18,690 --> 00:23:21,130
features just don't do. It seems to float around,

518
00:23:21,390 --> 00:23:24,390
spread across multiple vowels, stay behind when

519
00:23:24,390 --> 00:23:26,609
vowels get deleted. It behaves independently.

520
00:23:27,170 --> 00:23:30,490
So autosegmental phonology, ASP, developed by

521
00:23:30,490 --> 00:23:32,990
John Goldsmith and others, proposed what? It

522
00:23:32,990 --> 00:23:35,609
proposed that features like tone exist on a separate

523
00:23:35,609 --> 00:23:38,769
parallel tier, an auto -segmental tier, distinct

524
00:23:38,769 --> 00:23:40,710
from the tier where the consonants and vowels

525
00:23:40,710 --> 00:23:43,680
live. The segmental tier. Think of it like a

526
00:23:43,680 --> 00:23:45,940
musical melody line floating above the lyrics.

527
00:23:46,019 --> 00:23:47,819
And the melody only connects to the lyrics at

528
00:23:47,819 --> 00:23:50,380
certain points. Exactly. The tonal tier connects

529
00:23:50,380 --> 00:23:52,720
to the segmental tier via association lines,

530
00:23:52,859 --> 00:23:56,380
which link a tone, like H or L, to a specific

531
00:23:56,380 --> 00:23:59,619
tone -bearing unit, TBU, on the segmental tier,

532
00:23:59,799 --> 00:24:02,079
usually a vowel or mora. And this separation

533
00:24:02,079 --> 00:24:05,039
explains tone's weird behaviors. The source material

534
00:24:05,039 --> 00:24:08,339
listed five key ones. Yes. ASP provides elegant

535
00:24:08,339 --> 00:24:10,559
explanations for all of them. One, mobility.

536
00:24:10,960 --> 00:24:13,400
If a tone's association line can be detached

537
00:24:13,400 --> 00:24:16,059
and relinked to a different TBU, that explains

538
00:24:16,059 --> 00:24:18,960
why tones can seem to shift or dock onto neighboring

539
00:24:18,960 --> 00:24:21,420
syllables, like in chizigula. Features like plus

540
00:24:21,420 --> 00:24:24,559
nasal can't just jump ship like that. Two, stability.

541
00:24:24,960 --> 00:24:29,200
If the TBU, say a vowel, gets deleted, the tone

542
00:24:29,200 --> 00:24:30,799
that was linked to it doesn't have to disappear.

543
00:24:31,160 --> 00:24:33,940
it can remain on its tier and relink to a neighboring

544
00:24:33,940 --> 00:24:37,339
TVU that survives. This explains why tone often

545
00:24:37,339 --> 00:24:40,259
survives vowel deletion in languages like Haya

546
00:24:40,259 --> 00:24:42,819
or Pirahon. Right. The tone persists even when

547
00:24:42,819 --> 00:24:45,900
its original anchor is gone. Three. One -to -many

548
00:24:45,900 --> 00:24:48,900
association. A single tone on the tonal tier,

549
00:24:49,059 --> 00:24:52,180
say... one H, can have association lines linking

550
00:24:52,180 --> 00:24:55,319
it to multiple TBUs on the segmental tier. That's

551
00:24:55,319 --> 00:24:58,079
a tonal spreading or plateau where a whole sequence

552
00:24:58,079 --> 00:25:00,259
of syllables shares the same high pitch, like

553
00:25:00,259 --> 00:25:02,460
the unbounded H spread in Cholungo. Makes sense.

554
00:25:02,700 --> 00:25:05,880
Four, many -to -one association. This is the

555
00:25:05,880 --> 00:25:07,819
flip side. You can have multiple tones on the

556
00:25:07,819 --> 00:25:10,119
tonal tier, like an L followed by an H, both

557
00:25:10,119 --> 00:25:12,160
linking to a single TBU on the segmental tier.

558
00:25:12,460 --> 00:25:14,640
That's the autosegmental analysis of a contour

559
00:25:14,640 --> 00:25:17,180
tone. Multiple tones sharing one docking spot.

560
00:25:17,380 --> 00:25:19,849
Which links back to our contour discussion. Exactly.

561
00:25:20,069 --> 00:25:24,650
And five, toneless segments. ASP allows for TBUs

562
00:25:24,650 --> 00:25:27,029
on the segmental tier that simply don't have

563
00:25:27,029 --> 00:25:29,569
an association line linking them to any tone

564
00:25:29,569 --> 00:25:32,670
on the tonal tier, at least underlyingly. Their

565
00:25:32,670 --> 00:25:34,529
surface pitch is then filled in later by phonetic

566
00:25:34,529 --> 00:25:37,130
rules, like interpolation. Japanese pitch accent

567
00:25:37,130 --> 00:25:39,809
has syllables like this. That framework seems

568
00:25:39,809 --> 00:25:42,349
incredibly powerful for capturing how tone works.

569
00:25:42,890 --> 00:25:46,049
But ASP, at least initially, often assumed its

570
00:25:46,049 --> 00:25:48,789
rules and principles were universal and inviolable.

571
00:25:49,099 --> 00:25:52,240
That's where Actimality Theory, OT, came in,

572
00:25:52,259 --> 00:25:54,420
right? To explain the variation between languages.

573
00:25:54,839 --> 00:25:57,279
Precisely. OT, developed in the early 90s, took

574
00:25:57,279 --> 00:25:59,579
the insides of ASP, like tiers and association,

575
00:26:00,000 --> 00:26:02,779
but provided a new way to think about cross -linguistic

576
00:26:02,779 --> 00:26:05,299
variation. It said basically all languages share

577
00:26:05,299 --> 00:26:07,119
the same set of universal constraints, but they

578
00:26:07,119 --> 00:26:08,880
rank those constraints differently in terms of

579
00:26:08,880 --> 00:26:10,460
importance. Constraints. What kind of constraints

580
00:26:10,460 --> 00:26:12,099
are we talking about for tone? They fall into

581
00:26:12,099 --> 00:26:14,599
two broad categories. Facefulness constraints

582
00:26:14,599 --> 00:26:16,839
demand that the output form, what you actually

583
00:26:16,839 --> 00:26:19,269
say, be faithful to the input form the underlying

584
00:26:19,269 --> 00:26:23,410
representation for tone key ones are max st don't

585
00:26:23,410 --> 00:26:27,009
delete tones from the input and dpt don't insert

586
00:26:27,009 --> 00:26:30,210
new tones that weren't in the input okay so faithfulness

587
00:26:30,210 --> 00:26:32,930
says keep things the same what's the opposing

588
00:26:32,930 --> 00:26:36,710
force that's markedness constraints these demand

589
00:26:36,710 --> 00:26:40,210
that the output form be simple natural or unmarked

590
00:26:40,210 --> 00:26:43,150
for tone common markedness constraints include

591
00:26:43,150 --> 00:26:46,470
things like h High turns are articulatorily costly.

592
00:26:46,710 --> 00:26:50,269
Avoid them if possible. Or no contour tour. Avoid

593
00:26:50,269 --> 00:26:52,630
contour tones on a single TBU. They're complex.

594
00:26:52,910 --> 00:26:55,390
So language variation arises from the conflict

595
00:26:55,390 --> 00:26:57,750
between wanting to be faithful and wanting to

596
00:26:57,750 --> 00:27:00,230
be simple. Exactly. Every language grammar is

597
00:27:00,230 --> 00:27:03,069
essentially a specific ranking of these universal

598
00:27:03,069 --> 00:27:06,289
but viable constraints. The optimal output for

599
00:27:06,289 --> 00:27:08,289
given input is the one that violates the lowest

600
00:27:08,289 --> 00:27:10,670
ranked constraints possible. Let's take a concrete

601
00:27:10,670 --> 00:27:12,940
example. The source material heavily features

602
00:27:12,940 --> 00:27:16,140
the obligatory contour principle, OCP. What is

603
00:27:16,140 --> 00:27:18,319
that, and how does OT explain its different effects?

604
00:27:18,640 --> 00:27:21,420
The OCP is a powerful markedness constraint that

605
00:27:21,420 --> 00:27:24,240
basically says, adjacent identical elements are

606
00:27:24,240 --> 00:27:27,339
forbidden. In tonology, it typically applies

607
00:27:27,339 --> 00:27:29,539
to adjacent identical tones on the tonal tier,

608
00:27:29,680 --> 00:27:33,259
especially high tones, so HH is bad. So the OCP

609
00:27:33,259 --> 00:27:36,380
wants to get rid of sequences like HH. How can

610
00:27:36,380 --> 00:27:38,480
languages resolve that conflict differently based

611
00:27:38,480 --> 00:27:40,970
on constraint ranking? let's consider the conflict

612
00:27:40,970 --> 00:27:45,529
between ocp avoid hh and mxt keep all input tones

613
00:27:45,529 --> 00:27:47,589
we can see three different outcomes depending

614
00:27:47,589 --> 00:27:51,809
on how they're ranked one ocp maxt deletion If

615
00:27:51,809 --> 00:27:54,049
avoiding the OCP violation is more important

616
00:27:54,049 --> 00:27:56,329
than preserving every input tone, the language

617
00:27:56,329 --> 00:27:58,730
will simply delete one of the adjacent H tones

618
00:27:58,730 --> 00:28:01,950
and input HH surfaces as just H. This is exactly

619
00:28:01,950 --> 00:28:04,130
what happens in Mucin's rule. Famous in many

620
00:28:04,130 --> 00:28:07,349
Bantu languages like Shona, the OCP forces deletion.

621
00:28:07,470 --> 00:28:09,930
Okay, OCP wins, the tone gets deleted. What's

622
00:28:09,930 --> 00:28:14,430
another option? Two, OCP no fusion. Fusion. Maybe

623
00:28:14,430 --> 00:28:17,589
deleting a tone is bad, max T is high, but merging

624
00:28:17,589 --> 00:28:20,299
two adjacent tones into one is acceptable. No

625
00:28:20,299 --> 00:28:23,400
fusion is ranked lower than OCP. In this case,

626
00:28:23,539 --> 00:28:25,759
an input H and H might surface as a single H

627
00:28:25,759 --> 00:28:29,019
tone linked to both original TBUs. The OCP is

628
00:28:29,019 --> 00:28:31,039
satisfied because there's only one H node on

629
00:28:31,039 --> 00:28:34,019
the tier, but arguably both input tones are preserved

630
00:28:34,019 --> 00:28:36,839
in the linkage. So the HH sequence is avoided

631
00:28:36,839 --> 00:28:40,000
by merging them. What if OCP is ranked low? Three,

632
00:28:40,079 --> 00:28:44,130
MAXT, OCP. Violation tolerated. If preserving

633
00:28:44,130 --> 00:28:46,410
input tones is paramount and the language doesn't

634
00:28:46,410 --> 00:28:49,509
mind OCP violations much, then an input HH will

635
00:28:49,509 --> 00:28:53,160
simply surface as HH. The OCP constraint is violated,

636
00:28:53,420 --> 00:28:55,660
but that's deemed acceptable to satisfy the higher

637
00:28:55,660 --> 00:28:58,319
-ranked Max -T. Languages like Kishamba, another

638
00:28:58,319 --> 00:29:00,460
Bantu language, actually allow sequences of H

639
00:29:00,460 --> 00:29:02,420
-tones, often realized phonetically with down

640
00:29:02,420 --> 00:29:04,700
-step between them. The ranking dictates whether

641
00:29:04,700 --> 00:29:07,359
you delete, fuse, or tolerate the OCP violation.

642
00:29:07,779 --> 00:29:10,079
That OT framework really clarifies how languages

643
00:29:10,079 --> 00:29:11,980
can use the same basic ingredients but end up

644
00:29:11,980 --> 00:29:13,900
with very different surface patterns. Okay, moving

645
00:29:13,900 --> 00:29:16,039
into Section 4, let's see how tone interacts

646
00:29:16,039 --> 00:29:17,740
with the rest of the grammar. It's not just marking

647
00:29:17,740 --> 00:29:20,970
words, is it? Not at all. Tone is deeply integrated.

648
00:29:21,130 --> 00:29:23,750
It functions morphologically as a meaningful

649
00:29:23,750 --> 00:29:26,410
effix in its own right. Like adding a suffix,

650
00:29:26,410 --> 00:29:29,369
but it's just a tone. Exactly. We mentioned Cantonese

651
00:29:29,369 --> 00:29:32,049
using a specific high tone, often a changed tone

652
00:29:32,049 --> 00:29:35,009
derived from a high level 55, to mark familiar

653
00:29:35,009 --> 00:29:38,009
names or diminutives. You add the tone and the

654
00:29:38,009 --> 00:29:41,470
meaning changes. Or look at Aieo, a language

655
00:29:41,470 --> 00:29:44,750
from New Guinea. Verbal aspect, whether an action

656
00:29:44,750 --> 00:29:47,269
is ongoing, completed, etc., can be signaled

657
00:29:47,269 --> 00:29:49,769
simply by imposing a different fixed tonal melody

658
00:29:49,769 --> 00:29:52,809
onto the same monosyllabic verb root. The tone

659
00:29:52,809 --> 00:29:55,509
is the aspect morpheme. That's incredible. What

660
00:29:55,509 --> 00:29:57,890
about more complex morphological processes like

661
00:29:57,890 --> 00:30:00,430
reduplication? Does tone just get copied along

662
00:30:00,430 --> 00:30:01,849
with the segments? That's where it gets really

663
00:30:01,849 --> 00:30:04,160
interesting because the answer is... It depends.

664
00:30:04,579 --> 00:30:06,779
Reduplication acts like a diagnostic. In Mandarin

665
00:30:06,779 --> 00:30:08,880
Chinese, for example, how tone behaves under

666
00:30:08,880 --> 00:30:11,059
reduplication depends entirely on the grammatical

667
00:30:11,059 --> 00:30:14,099
category of the word being reduplicated. If you

668
00:30:14,099 --> 00:30:17,460
reduplicate certain kinship nouns, the tone on

669
00:30:17,460 --> 00:30:19,700
the reduplicated part is deleted, leaving it

670
00:30:19,700 --> 00:30:23,200
toneless. If you reduplicate classifiers, measure

671
00:30:23,200 --> 00:30:25,940
words, the original tone is copied faithfully

672
00:30:25,940 --> 00:30:29,019
onto the reduplicated part, tone tone. But if

673
00:30:29,019 --> 00:30:31,759
you reduplicate adjectives to form adverbs, the

674
00:30:31,759 --> 00:30:34,440
reduplicated part gets a fixed high tone, 55,

675
00:30:34,779 --> 00:30:36,519
regardless of the adjective's original tone,

676
00:30:36,660 --> 00:30:39,859
tone 55. Wow. So the grammar dictates whether

677
00:30:39,859 --> 00:30:42,140
the tone is deleted, copied, or replaced based

678
00:30:42,140 --> 00:30:44,400
on the construction. Precisely. It shows that

679
00:30:44,400 --> 00:30:47,180
tonal rules aren't just blind phonological processes.

680
00:30:47,400 --> 00:30:50,039
They're sensitive to morphological and syntactic

681
00:30:50,039 --> 00:30:52,960
context. This gets even more intricate in the

682
00:30:52,960 --> 00:30:55,220
complex verb structures of many Bantu languages.

683
00:30:55,579 --> 00:30:57,880
Ah, the morphological domains. What's going on

684
00:30:57,880 --> 00:31:00,130
there? Bantu verbs are famous for having multiple

685
00:31:00,130 --> 00:31:02,710
suffixes added to a root, creating very long

686
00:31:02,710 --> 00:31:05,710
words. Crucially, this morphology defines specific

687
00:31:05,710 --> 00:31:08,549
nested domains within the verb. There's the basic

688
00:31:08,549 --> 00:31:11,609
verb stem, then a larger verbal macro stem, including

689
00:31:11,609 --> 00:31:14,430
extensions, then an even larger inflectional

690
00:31:14,430 --> 00:31:17,490
stem, including tense suspect markers. And tonal

691
00:31:17,490 --> 00:31:20,190
rules care about these domains. Absolutely. Rules

692
00:31:20,190 --> 00:31:23,250
like the OCP -driven processes, like Meason's

693
00:31:23,250 --> 00:31:26,269
rule, deleting the second H and HH, often apply

694
00:31:26,269 --> 00:31:28,609
cyclically within these domains. A rule might

695
00:31:28,609 --> 00:31:31,190
apply first within the stem, then its output

696
00:31:31,190 --> 00:31:33,230
becomes the input for the rule applying across

697
00:31:33,230 --> 00:31:36,210
the macro stem and so on. The phonology has to

698
00:31:36,210 --> 00:31:39,250
precisely track the internal morphological boundaries.

699
00:31:39,549 --> 00:31:41,809
Okay, shifting from word internal stuff to the

700
00:31:41,809 --> 00:31:44,349
sentence level. How do tone languages handle

701
00:31:44,349 --> 00:31:46,730
things like asking questions, showing surprise,

702
00:31:47,069 --> 00:31:50,069
or indicating topic -focused things that non

703
00:31:50,069 --> 00:31:52,410
-tonal languages often do with overall sentence

704
00:31:52,410 --> 00:31:55,109
intonation? That's a great question. While some

705
00:31:55,109 --> 00:31:57,470
tone languages do also use overall pitch range

706
00:31:57,470 --> 00:32:00,130
modifications for pragmatic effect, like raising

707
00:32:00,130 --> 00:32:02,170
the whole pitch register for a question in dang

708
00:32:02,170 --> 00:32:05,349
may, many rely heavily on sentence final particles.

709
00:32:05,609 --> 00:32:07,230
The little words tacked on to the end of the

710
00:32:07,230 --> 00:32:09,650
sentence. Exactly. These particles often carry

711
00:32:09,650 --> 00:32:11,829
little segmental meaning themselves, but they

712
00:32:11,829 --> 00:32:13,950
bear specific tones that convey the pragmatic

713
00:32:13,950 --> 00:32:17,130
or illocutionary force. Mandarin Chinese has

714
00:32:17,130 --> 00:32:20,009
several particles like ma for questions, ba for

715
00:32:20,009 --> 00:32:23,150
suggestions, ne for continuation. And Cantonese

716
00:32:23,150 --> 00:32:25,549
is even more extreme. Cantonese is famous for

717
00:32:25,549 --> 00:32:28,109
it, reportedly having somewhere between 35 and

718
00:32:28,109 --> 00:32:31,430
40 distinct sentence final particles. They often

719
00:32:31,430 --> 00:32:33,670
use a segmentally minimal particle, like just

720
00:32:33,670 --> 00:32:35,930
a vowel A, which then serves as a host for different

721
00:32:35,930 --> 00:32:38,890
tones to signal very subtle pragmatic nuances,

722
00:32:39,410 --> 00:32:42,150
doubt, confirmation, impatience, shared knowledge,

723
00:32:42,289 --> 00:32:45,109
and so on. The particle is almost just a placeholder

724
00:32:45,109 --> 00:32:48,029
for the pragmatic tone. So the pragmatics gets

725
00:32:48,029 --> 00:32:50,910
segmentalized onto these particles. How does

726
00:32:50,910 --> 00:32:53,109
the larger syntactic structure, the sentence

727
00:32:53,109 --> 00:32:56,589
tree, influence tone syntax plays a crucial role

728
00:32:56,589 --> 00:32:59,910
in defining the phonological phrase phvh the

729
00:32:59,910 --> 00:33:02,450
phvh is the typical domain within which many

730
00:33:02,450 --> 00:33:05,109
tonal rules especially sandy rules where tones

731
00:33:05,109 --> 00:33:07,869
change based on context apply and how is the

732
00:33:07,869 --> 00:33:10,430
phvh determined it's often determined by mapping

733
00:33:10,430 --> 00:33:12,650
rules that look at the syntactic structure a

734
00:33:12,650 --> 00:33:15,230
common mechanism is edge alignment Languages

735
00:33:15,230 --> 00:33:17,630
differ on which edge matters. For instance, in

736
00:33:17,630 --> 00:33:20,269
Xiamen Chinese, a min dialect, the right edge

737
00:33:20,269 --> 00:33:22,549
of most syntactic phrases, like verb phrases

738
00:33:22,549 --> 00:33:25,170
or noun phrases excluding adjuncts, must align

739
00:33:25,170 --> 00:33:26,769
with the boundary of a tone group, triggering

740
00:33:26,769 --> 00:33:29,410
specific Sandy rules within that group. Other

741
00:33:29,410 --> 00:33:31,509
languages might align rules to the left edge

742
00:33:31,509 --> 00:33:34,009
of phrases. So the syntax carves up the sentence

743
00:33:34,009 --> 00:33:37,680
into domains for tone rules. What about Mandarin's

744
00:33:37,680 --> 00:33:41,049
famous third tone, Sandy? Where L plus L becomes

745
00:33:41,049 --> 00:33:43,890
LH plus L. Does that respect syntactic boundaries?

746
00:33:44,329 --> 00:33:46,930
That's where it gets even more interesting. Mandarin

747
00:33:46,930 --> 00:33:50,490
tone zandy is famously complex. While sensitive

748
00:33:50,490 --> 00:33:53,490
to syntax to some extent, it also shows a strong

749
00:33:53,490 --> 00:33:55,990
preference for creating binary prosodic constituents,

750
00:33:56,230 --> 00:33:59,170
roughly. Two -syllable rhythmic units, or feet,

751
00:33:59,349 --> 00:34:01,730
sometimes called metrical rhythmic units, MRUs.

752
00:34:02,029 --> 00:34:04,250
So the rhythm matters more than the syntax sometimes.

753
00:34:04,650 --> 00:34:07,599
It seems so. The third -tone Sandy rule applies

754
00:34:07,599 --> 00:34:09,900
cyclically, often creating these binary feet.

755
00:34:10,119 --> 00:34:12,659
It might apply within a two -syllable foot, even

756
00:34:12,659 --> 00:34:15,380
if that foot crosses a syntactic boundary, or

757
00:34:15,380 --> 00:34:17,380
it might fail to apply between two third -tones

758
00:34:17,380 --> 00:34:19,139
if they belong to different feet, even if they're

759
00:34:19,139 --> 00:34:21,820
adjacent syntactically. The prosodic structure,

760
00:34:22,159 --> 00:34:24,320
the rhythmic organization, seems to exert its

761
00:34:24,320 --> 00:34:27,800
own influence, sometimes overwriting strict syntactic

762
00:34:27,800 --> 00:34:30,079
domains. That adds another layer of complexity.

763
00:34:30,679 --> 00:34:32,739
Okay, let's put these theoretical tools into

764
00:34:32,739 --> 00:34:35,119
action now, looking at Section 5 regional patterns.

765
00:34:35,260 --> 00:34:37,599
We'll start with African systems, which you mentioned

766
00:34:37,599 --> 00:34:40,780
are characterized by mobility and downstep. Yes,

767
00:34:40,780 --> 00:34:43,539
if you had to pick one defining feature of many

768
00:34:43,539 --> 00:34:46,500
African tone systems, especially Bantu, it's

769
00:34:46,500 --> 00:34:50,219
tonal mobility. Unlike some Asian languages where

770
00:34:50,219 --> 00:34:52,800
a tone feels permanently welded to its lexical

771
00:34:52,800 --> 00:34:55,840
root, African tones often seem much more independent.

772
00:34:56,280 --> 00:34:59,420
How does that mobility manifest? Their surface

773
00:34:59,420 --> 00:35:01,800
alignment, where the tone actually docks, seems

774
00:35:01,800 --> 00:35:04,280
less determined by strict lexical pre -linking

775
00:35:04,280 --> 00:35:07,659
and more by general phonological principles operating

776
00:35:07,659 --> 00:35:11,159
across phrases or words. Things like align a

777
00:35:11,159 --> 00:35:13,539
high tone to the stressed syllable or align a

778
00:35:13,539 --> 00:35:16,380
tone to the left ray edge of the word or shift

779
00:35:16,380 --> 00:35:19,539
a tone to avoid an OCP violation. The tones seem

780
00:35:19,539 --> 00:35:22,079
to float more readily and dock according to grammatical

781
00:35:22,079 --> 00:35:25,320
context. The example of jishang. from Cameroon

782
00:35:25,320 --> 00:35:27,000
was mentioned regarding the tone -bearing unit.

783
00:35:27,139 --> 00:35:29,159
Right. Zhishang provides really clear evidence

784
00:35:29,159 --> 00:35:32,199
that the mora is the TBU in that language. Crucially,

785
00:35:32,320 --> 00:35:35,039
contour tones, which need two timing units, are

786
00:35:35,039 --> 00:35:36,920
generally only allowed on syllables that are

787
00:35:36,920 --> 00:35:39,340
demonstrably bimaraic, either having a long vowel

788
00:35:39,340 --> 00:35:42,019
or a final consonant that contributes moraic

789
00:35:42,019 --> 00:35:43,960
weight. And it also showed something about which

790
00:35:43,960 --> 00:35:47,440
sounds can actually carry the pitch. Yes. Zhisheng

791
00:35:47,440 --> 00:35:49,480
Phonetics reveals that only sonorant segments,

792
00:35:49,880 --> 00:35:52,780
vowels, nasals, liquids can be the physical bearers

793
00:35:52,780 --> 00:35:55,980
of the philodollar contour. Obstruents, stops

794
00:35:55,980 --> 00:35:58,460
fricatives, can contribute to syllable weight,

795
00:35:58,659 --> 00:36:01,280
be moraic, and thus provide a potential docking

796
00:36:01,280 --> 00:36:04,280
site for a tone phonologically, but they cannot

797
00:36:04,280 --> 00:36:06,840
themselves carry the audible pitch. The tone

798
00:36:06,840 --> 00:36:10,199
has to be realized on an adjacent sonorant. Fascinating

799
00:36:10,199 --> 00:36:12,300
phonetic constraint. Now the downstep controversy.

800
00:36:12,760 --> 00:36:15,820
Is it a real L -tone? causing the lowering downdrift

801
00:36:15,820 --> 00:36:19,059
or a ghost L floating tone? You said Africa is

802
00:36:19,059 --> 00:36:22,099
key here. It's a long standing debate. The unification

803
00:36:22,099 --> 00:36:24,599
approach tries to argue that all downstep ultimately

804
00:36:24,599 --> 00:36:26,900
derives from a low tone, whether it's overtly

805
00:36:26,900 --> 00:36:29,099
present or just historically deleted, but still

806
00:36:29,099 --> 00:36:31,719
active, essentially making all downstep a form

807
00:36:31,719 --> 00:36:34,159
of downdrift. Josham challenges that. Zhisheng

808
00:36:34,159 --> 00:36:35,820
seems to be a counterexample. According to the

809
00:36:35,820 --> 00:36:38,619
sources, it clearly exhibits non -automatic downstep

810
00:36:38,619 --> 00:36:41,159
high tones lowering without an overt preceding

811
00:36:41,159 --> 00:36:43,460
L, but it lacks automatic downdrift, where an

812
00:36:43,460 --> 00:36:46,179
overt L always triggers lowering. So you can

813
00:36:46,179 --> 00:36:48,599
have the ghost L effect without the real L effect.

814
00:36:49,039 --> 00:36:51,719
Apparently so, in Zhang. This suggests that the

815
00:36:51,719 --> 00:36:54,639
abstract floating L analysis, non -automatic

816
00:36:54,639 --> 00:36:57,099
downstep, might be necessary and distinct from

817
00:36:57,099 --> 00:36:59,460
simple downdrift, challenging theories that try

818
00:36:59,460 --> 00:37:02,150
to reduce everything to just one mechanism. The

819
00:37:02,150 --> 00:37:04,250
nature of downstep remains a complex research

820
00:37:04,250 --> 00:37:06,849
area. Okay, let's hop continents to Asian systems.

821
00:37:07,090 --> 00:37:09,550
The sources highlight their tonal richness and

822
00:37:09,550 --> 00:37:12,289
complex sandy chains. Yeah, if African systems

823
00:37:12,289 --> 00:37:15,010
often showcase mobility, many East and Southeast

824
00:37:15,010 --> 00:37:18,150
Asian systems showcase sheer inventory richness

825
00:37:18,150 --> 00:37:22,170
and intricate tone sandy. Contextual tone changes.

826
00:37:22,429 --> 00:37:24,349
Richness in what sense? The number of lexical

827
00:37:24,349 --> 00:37:27,139
tone contrasts can be very high. Cantonese famously

828
00:37:27,139 --> 00:37:29,599
has six or seven distinct tones, depending on

829
00:37:29,599 --> 00:37:32,260
the analysis. This tonal richness allows languages

830
00:37:32,260 --> 00:37:34,739
with relatively simple syllable structures, like

831
00:37:34,739 --> 00:37:37,619
Mandarin, which only has about 400 unique combinations

832
00:37:37,619 --> 00:37:40,840
of consonants and vowels, to create a vast lexicon.

833
00:37:41,159 --> 00:37:44,219
Right, you mentioned Mandarin goes from 406 segmental

834
00:37:44,219 --> 00:37:46,960
syllables to over 1 ,200 distinct syllables once

835
00:37:46,960 --> 00:37:49,800
you add the four main tones. Exactly. Tone dramatically

836
00:37:49,800 --> 00:37:52,480
expands the meaning -carrying capacity of a simple

837
00:37:52,480 --> 00:37:55,159
syllable inventory. Cantonese also has that change

838
00:37:55,159 --> 00:37:58,159
tone. Use morphologically. Yes, the change tone

839
00:37:58,159 --> 00:38:00,260
is interesting. It's often a high -rising or

840
00:38:00,260 --> 00:38:03,420
high -level tone, like 35 or 55, that replaces

841
00:38:03,420 --> 00:38:05,940
the original lexical tone, typically used for

842
00:38:05,940 --> 00:38:08,239
familiar names, diminutives, or certain types

843
00:38:08,239 --> 00:38:11,039
of common nouns. What's telling is that it often

844
00:38:11,039 --> 00:38:13,699
simplifies the original tone. If the original

845
00:38:13,699 --> 00:38:16,420
tone was a complex contour, like low -falling

846
00:38:16,420 --> 00:38:19,360
21, the change tone replaces it with a simpler

847
00:38:19,360 --> 00:38:22,320
high pitch. It suggests a morphological process

848
00:38:22,320 --> 00:38:25,219
overriding phonological faithfulness to the base

849
00:38:25,219 --> 00:38:28,739
tone. bending example of Sandy seems to be the

850
00:38:28,739 --> 00:38:31,139
min -chain shift in dialects like Shenanmen and

851
00:38:31,139 --> 00:38:33,360
Chaoyang. What on earth is going on there? It's

852
00:38:33,360 --> 00:38:35,619
truly remarkable and frankly sounds incredibly

853
00:38:35,619 --> 00:38:38,280
difficult to learn. It's an extreme form of tone

854
00:38:38,280 --> 00:38:41,420
Sandy where basically every tone in the language's

855
00:38:41,420 --> 00:38:43,619
inventory changes into a different tone from

856
00:38:43,619 --> 00:38:46,280
the inventory when it occurs in a non -final

857
00:38:46,280 --> 00:38:48,460
position within a phonological phrase. Every

858
00:38:48,460 --> 00:38:50,980
single one. It's not just one tone changing,

859
00:38:51,039 --> 00:38:53,699
but all of them shift. Pretty much. It creates

860
00:38:53,699 --> 00:38:56,039
a chain effect. For example, in Schemen, the

861
00:38:56,039 --> 00:38:58,900
citation tone 53, high falling, shifts to 31,

862
00:38:59,119 --> 00:39:02,320
low falling, in sandy context. The citation tone

863
00:39:02,320 --> 00:39:06,000
31, low falling, shifts to 55, high level. The

864
00:39:06,000 --> 00:39:09,239
citation 55 shifts elsewhere, and so on. No tone

865
00:39:09,239 --> 00:39:12,000
stays the same, and no two citation tones shift

866
00:39:12,000 --> 00:39:14,539
to the same sandy tone. That sounds less like

867
00:39:14,539 --> 00:39:16,480
a natural phonetic process, like assimilation,

868
00:39:16,699 --> 00:39:20,139
and more like just memorizing a giant substitution

869
00:39:20,139 --> 00:39:22,599
table. That's largely the consensus that these

870
00:39:22,599 --> 00:39:24,559
change shifts are significantly lexicalized.

871
00:39:24,920 --> 00:39:26,880
Learners likely have to memorize the specific

872
00:39:26,880 --> 00:39:29,960
pairing for each tone. Tone X becomes tone Y

873
00:39:29,960 --> 00:39:32,400
in this context. But is it completely random

874
00:39:32,400 --> 00:39:34,980
or is there some underlying logic? It's not entirely

875
00:39:34,980 --> 00:39:36,920
random. The pattern is constrained by principles

876
00:39:36,920 --> 00:39:40,219
of markedness and inventory reduction. Non -final

877
00:39:40,219 --> 00:39:42,480
positions in a phrase are typically less prominent,

878
00:39:42,619 --> 00:39:45,239
weaker positions. Languages often restrict the

879
00:39:45,239 --> 00:39:47,039
types of contrasts allowed in weak positions,

880
00:39:47,260 --> 00:39:50,320
for example. Complex contours or very high tones.

881
00:39:50,510 --> 00:39:52,809
be disallowed there ah so the target inventory

882
00:39:52,809 --> 00:39:55,489
for the sandy tone is smaller or simpler exactly

883
00:39:55,489 --> 00:39:58,389
the input tone has to shift but it can only shift

884
00:39:58,389 --> 00:40:00,329
to one of the tones that are permitted in that

885
00:40:00,329 --> 00:40:03,610
non -final weaker context it usually shifts to

886
00:40:03,610 --> 00:40:06,289
the closest available permissible tone because

887
00:40:06,289 --> 00:40:08,949
multiple input tones might ideally want to shift

888
00:40:08,949 --> 00:40:11,570
to the same simple output tone like mid -level

889
00:40:11,570 --> 00:40:15,219
but can't To maintain contrast, they're forced

890
00:40:15,219 --> 00:40:17,280
to shift elsewhere, creating the chain reaction.

891
00:40:17,579 --> 00:40:20,179
So it's memorized pairings. But the pairings

892
00:40:20,179 --> 00:40:22,500
themselves are shaped by phonological constraints

893
00:40:22,500 --> 00:40:25,679
on permissible tones in context. Wow. Okay, third

894
00:40:25,679 --> 00:40:29,000
continent, American systems. The sources point

895
00:40:29,000 --> 00:40:31,199
to stress tone interactions and floating tones,

896
00:40:31,380 --> 00:40:34,579
especially in mixtec. Yes, the Otomangian languages

897
00:40:34,579 --> 00:40:37,360
of Mexico, like Mixtec and Trach, are famous

898
00:40:37,360 --> 00:40:39,719
for their complex tonal systems, often involving

899
00:40:39,719 --> 00:40:42,000
intricate interactions with stress and morphology.

900
00:40:42,239 --> 00:40:44,460
Mixtec provides classic examples of floating

901
00:40:44,460 --> 00:40:46,559
tones behaving in very specific ways. Like the

902
00:40:46,559 --> 00:40:48,699
floating high tone in San Miguel Grande Mixtec.

903
00:40:48,800 --> 00:40:51,920
What does it do? This floating H tone is underlyingly

904
00:40:51,920 --> 00:40:54,619
present but not linked. It wants to dock and

905
00:40:54,619 --> 00:40:56,820
it moves rightward across the word looking for

906
00:40:56,820 --> 00:41:00,079
a place. Crucially, it tends to skip over vowels

907
00:41:00,079 --> 00:41:02,780
that are specified as toneless. which surfaces

908
00:41:02,780 --> 00:41:05,880
mid why skip them the analysis suggests it's

909
00:41:05,880 --> 00:41:08,179
governed by a constraint called tonal prominence

910
00:41:08,179 --> 00:41:11,059
faith basically the language wants to preserve

911
00:41:11,059 --> 00:41:13,460
the inherent tonal specification of each vowel

912
00:41:13,460 --> 00:41:16,579
landing the floating h on a toneless vowel would

913
00:41:16,579 --> 00:41:20,519
change its status from neutral mid to high violating

914
00:41:20,519 --> 00:41:23,619
this faithfulness so the h keeps moving until

915
00:41:23,619 --> 00:41:25,780
it finds a vowel that already has a lexical tone

916
00:41:25,780 --> 00:41:29,079
h or l where docking is less disruptive to the

917
00:41:29,079 --> 00:41:32,199
overall profile That's incredibly specific behavior.

918
00:41:32,400 --> 00:41:35,500
Does anything stop this floating H? Yes. Glottal

919
00:41:35,500 --> 00:41:38,119
stops are reported to act as barriers. The floating

920
00:41:38,119 --> 00:41:40,400
H cannot cross or dock onto a vowel immediately

921
00:41:40,400 --> 00:41:43,460
following a glottal stop. It shows that segmental

922
00:41:43,460 --> 00:41:46,139
features can interact with and block purely tonal

923
00:41:46,139 --> 00:41:49,280
processes. Mixtech also flips the usual stress

924
00:41:49,280 --> 00:41:51,139
-tone relationship, right? Usually stress attracts

925
00:41:51,139 --> 00:41:54,059
tone, but here... Here, tone seems to determine

926
00:41:54,059 --> 00:41:57,309
stress. In some mixed tech varieties, stress

927
00:41:57,309 --> 00:42:00,269
placement isn't fixed lexically, but is assigned

928
00:42:00,269 --> 00:42:03,489
based on the tonal sequence. Specifically, stress

929
00:42:03,489 --> 00:42:05,650
tends to fall on a syllable that is immediately

930
00:42:05,650 --> 00:42:07,730
followed by a syllable bearing a lower pitch.

931
00:42:08,030 --> 00:42:11,550
So like H followed by M or L, or M followed by

932
00:42:11,550 --> 00:42:14,389
L. Exactly. The pitch drop seems to create the

933
00:42:14,389 --> 00:42:16,809
prominence. Stress falls on the peak right before

934
00:42:16,809 --> 00:42:19,269
the descent. It's tone -dependent stress, which

935
00:42:19,269 --> 00:42:22,070
is typologically quite rare. And finally for

936
00:42:22,070 --> 00:42:24,989
the Americas, TRIQ shows tone acting as pure

937
00:42:24,989 --> 00:42:27,389
morphology. Completely. TRIQ has grammatical

938
00:42:27,389 --> 00:42:30,010
morphemes, particularly person and number markers,

939
00:42:30,150 --> 00:42:32,670
on verbs that consist solely of a floating tone.

940
00:42:32,909 --> 00:42:35,449
For instance, adding a floating four tone on

941
00:42:35,449 --> 00:42:37,610
a one to five scale might mark the first person

942
00:42:37,610 --> 00:42:41,059
plural exclusive. This floating tone docks onto

943
00:42:41,059 --> 00:42:43,400
the verb root, combining with the root's own

944
00:42:43,400 --> 00:42:46,139
tone to create a specific contour tone that signals

945
00:42:46,139 --> 00:42:48,159
that grammatical meaning. No segments at all,

946
00:42:48,179 --> 00:42:49,980
just the tone carries the grammatical function.

947
00:42:50,219 --> 00:42:53,019
Amazing. Okay, final section, section six, perception

948
00:42:53,019 --> 00:42:55,960
and acquisition. How do our brains actually handle

949
00:42:55,960 --> 00:42:59,019
these complex systems? Starting with adult perception.

950
00:42:59,320 --> 00:43:01,500
Our ears are pretty sensitive, aren't they? Incredibly

951
00:43:01,500 --> 00:43:04,739
sensitive. Psychoacoustic tests show we can detect

952
00:43:04,739 --> 00:43:07,900
just noticeable differences, JNDs, in frequency

953
00:43:07,900 --> 00:43:11,519
as small as 0 .3 hertz in ideal conditions. But

954
00:43:11,519 --> 00:43:13,800
languages don't use contrast that small, do they?

955
00:43:13,920 --> 00:43:16,420
Not even close. Even languages with relatively

956
00:43:16,420 --> 00:43:19,719
crowded tonal spaces, like Mambila with four

957
00:43:19,719 --> 00:43:22,599
level tones, keep those tones spaced much further

958
00:43:22,599 --> 00:43:25,260
apart, maybe 10 hertz or more between adjacent

959
00:43:25,260 --> 00:43:28,019
levels. Why the big gap? It ensures robustness.

960
00:43:28,179 --> 00:43:30,780
Languages build in a large perceptual buffer

961
00:43:30,780 --> 00:43:33,460
zone so that minor production variations or noise

962
00:43:33,460 --> 00:43:35,900
don't cause confusion between phonologically

963
00:43:35,900 --> 00:43:38,579
distinct categories. We can hear tiny differences,

964
00:43:38,800 --> 00:43:41,739
but languages only make use of much larger, reliably

965
00:43:41,739 --> 00:43:44,360
perceivable differences for contrast. And there

966
00:43:44,360 --> 00:43:46,699
seems to be a cognitive limit on how many levels

967
00:43:46,699 --> 00:43:49,139
we can reliably track. The five -tone maximum.

968
00:43:49,630 --> 00:43:52,170
Yes, experimental work by Pollack back in the

969
00:43:52,170 --> 00:43:54,929
50s asking listeners to identify absolute pitch

970
00:43:54,929 --> 00:43:57,630
levels presented in isolation found that performance

971
00:43:57,630 --> 00:43:59,769
drops off sharply beyond about five categories.

972
00:43:59,889 --> 00:44:02,250
People just can't reliably assign sounds to more

973
00:44:02,250 --> 00:44:04,409
than five absolute pitch bins. And this might

974
00:44:04,409 --> 00:44:06,590
explain why no language seems to have more than

975
00:44:06,590 --> 00:44:09,449
five contrastive level tones. It's considered

976
00:44:09,449 --> 00:44:12,190
a likely perceptual constraint. While we can

977
00:44:12,190 --> 00:44:15,219
hear relative pitch changes very well, Our ability

978
00:44:15,219 --> 00:44:17,920
to identify absolute pitch levels is limited,

979
00:44:18,000 --> 00:44:20,900
and this seems to cap the maximum inventory size

980
00:44:20,900 --> 00:44:23,320
for level tones globally. We also have evidence

981
00:44:23,320 --> 00:44:25,739
that our brains treat tone linguistically, not

982
00:44:25,739 --> 00:44:28,360
just as raw sound, right? Yeah. The idea of categorical

983
00:44:28,360 --> 00:44:30,920
perception. Absolutely. Classic studies compared

984
00:44:30,920 --> 00:44:33,000
native Mandarin speakers and English speakers

985
00:44:33,000 --> 00:44:35,320
listening to synthesized sounds that gradually

986
00:44:35,320 --> 00:44:38,320
changed from a level tone to a rising tone. What

987
00:44:38,320 --> 00:44:40,869
was the difference? The English speakers perceived

988
00:44:40,869 --> 00:44:43,809
a smooth, continuous change. As the rise got

989
00:44:43,809 --> 00:44:47,170
steeper, they just heard it as more rising. The

990
00:44:47,170 --> 00:44:49,210
Mandarin speakers, however, perceived a sharp,

991
00:44:49,289 --> 00:44:52,489
sudden switch. Sounds on one side of a specific

992
00:44:52,489 --> 00:44:55,090
acoustic boundary around a 12 hertz rise were

993
00:44:55,090 --> 00:44:57,909
consistently labeled level tone, and sounds just

994
00:44:57,909 --> 00:44:59,789
slightly different on the other side were consistently

995
00:44:59,789 --> 00:45:02,590
labeled rising tone. So their language experience

996
00:45:02,590 --> 00:45:04,730
created a category boundary where physically

997
00:45:04,730 --> 00:45:08,000
there was none. Exactly. They perceive the sounds

998
00:45:08,000 --> 00:45:10,039
through the filter of their phonological system.

999
00:45:10,500 --> 00:45:13,320
Pitch perception becomes categorical for linguistically

1000
00:45:13,320 --> 00:45:16,039
relevant contrasts, just like how English speakers

1001
00:45:16,039 --> 00:45:18,760
perceive a sharp boundary between B and P along

1002
00:45:18,760 --> 00:45:21,340
the voice onset time continuum. What helps us

1003
00:45:21,340 --> 00:45:24,599
perceive tone besides the raw pitch, Fever Dollars?

1004
00:45:24,619 --> 00:45:26,340
Well, Fever Dollars is definitely the primary

1005
00:45:26,340 --> 00:45:29,269
cue. Other things help. Amplitude, loudness,

1006
00:45:29,309 --> 00:45:31,690
often mirrors the Fevedalar contour higher pitch,

1007
00:45:31,849 --> 00:45:34,030
often correlates with slightly higher amplitude,

1008
00:45:34,269 --> 00:45:37,389
and this can provide a secondary cue. Also, as

1009
00:45:37,389 --> 00:45:39,630
we mentioned, listeners perform incredible feats

1010
00:45:39,630 --> 00:45:42,309
of contextual compensation, normalizing for the

1011
00:45:42,309 --> 00:45:44,309
speaker's overall pitch range and accounting

1012
00:45:44,309 --> 00:45:47,130
for effects like declination and downdrift to

1013
00:45:47,130 --> 00:45:49,369
identify the intended underlying tone correctly.

1014
00:45:49,710 --> 00:45:51,829
Okay, let's turn to kids. First language acquisition,

1015
00:45:52,230 --> 00:45:55,489
L1. How early do infants tune into the tones

1016
00:45:55,489 --> 00:45:58,199
of their language? amazingly early. Studies with

1017
00:45:58,199 --> 00:46:00,500
infants exposed to tonal languages like Yoruba

1018
00:46:00,500 --> 00:46:02,780
show that even by six to eight months old, they

1019
00:46:02,780 --> 00:46:04,579
are already better at discriminating a specific

1020
00:46:04,579 --> 00:46:06,719
tonal contrast relevant to their language, like

1021
00:46:06,719 --> 00:46:09,340
high versus non -high, compared to infants learning

1022
00:46:09,340 --> 00:46:11,539
non -tonal languages like English. So they're

1023
00:46:11,539 --> 00:46:13,679
already specializing perceptually before they

1024
00:46:13,679 --> 00:46:16,639
can even speak. Yes. And conversely, English

1025
00:46:16,639 --> 00:46:19,000
-learning infants, who might initially be sensitive

1026
00:46:19,000 --> 00:46:22,239
to some tonal contrasts, gradually lose that

1027
00:46:22,239 --> 00:46:24,699
sensitivity because it's not relevant or contrastive

1028
00:46:24,699 --> 00:46:26,960
in their linguistic input. The environment shapes

1029
00:46:26,960 --> 00:46:29,619
the perceptual system very early on. What about

1030
00:46:29,619 --> 00:46:31,960
when kids start producing tones? Is there a universal

1031
00:46:31,960 --> 00:46:33,960
order, like certain tones are always easier?

1032
00:46:34,139 --> 00:46:36,260
Not entirely universal. It's heavily influenced

1033
00:46:36,260 --> 00:46:38,940
by the specific target language system. While

1034
00:46:38,940 --> 00:46:41,059
there's a general tendency for rising tones,

1035
00:46:41,320 --> 00:46:44,360
L plus H, to be acquired later across languages,

1036
00:46:44,500 --> 00:46:47,099
likely due to their articulatory complexity.

1037
00:46:47,480 --> 00:46:49,179
Right, the production difficulty we talked about.

1038
00:46:49,360 --> 00:46:52,329
The specific order for level tones varies. In

1039
00:46:52,329 --> 00:46:54,570
Mandarin, kids typically master the high -level

1040
00:46:54,570 --> 00:46:58,309
55 and high -falling 53 tones first. The dipping

1041
00:46:58,309 --> 00:47:02,150
214 and rising 35 tones come much later and are

1042
00:47:02,150 --> 00:47:05,510
more error -prone. In Cantonese, it's the extreme

1043
00:47:05,510 --> 00:47:08,090
-level tones the highest 55 and the lowest 11

1044
00:47:08,090 --> 00:47:10,750
that are acquired earliest. In Thai, interestingly,

1045
00:47:10,989 --> 00:47:13,389
the mid -level and low -level tones tend to be

1046
00:47:13,389 --> 00:47:15,889
acquired first, with high and falling tones coming

1047
00:47:15,889 --> 00:47:18,750
later. So kids prioritize the most distinct or

1048
00:47:18,750 --> 00:47:21,369
perhaps most frequent tones in their specific

1049
00:47:21,369 --> 00:47:24,070
input language. It seems that way. They latch

1050
00:47:24,070 --> 00:47:26,730
onto the most stable and contrastive points in

1051
00:47:26,730 --> 00:47:29,909
their particular system first. And finally, what

1052
00:47:29,909 --> 00:47:32,289
about acquiring the complex rules involving tone?

1053
00:47:32,699 --> 00:47:35,480
like Sandy or the rules linked to syntax. Those

1054
00:47:35,480 --> 00:47:37,480
tonal alternations generally come relatively

1055
00:47:37,480 --> 00:47:40,659
late in acquisition, often well after children

1056
00:47:40,659 --> 00:47:43,599
have mastered basic sentence structure and segmental

1057
00:47:43,599 --> 00:47:46,579
phonology. Mandarin third tone Sandy, for instance,

1058
00:47:46,820 --> 00:47:49,599
is notoriously difficult, with kids still making

1059
00:47:49,599 --> 00:47:52,559
frequent errors even after age three. And Deborah

1060
00:47:52,559 --> 00:47:54,860
DeMuth's work on Sosotho showed a difference

1061
00:47:54,860 --> 00:47:57,760
between purely phonological rules and syntax

1062
00:47:57,760 --> 00:48:00,280
phonology interface rules. Yes, that's a crucial

1063
00:48:00,280 --> 00:48:03,280
finding. Sisotho children seem to master purely

1064
00:48:03,280 --> 00:48:06,280
phonological tonal rules, like high tone doubling,

1065
00:48:06,519 --> 00:48:08,659
spreading an H tone, relatively early, maybe

1066
00:48:08,659 --> 00:48:10,980
around age three. But rules whose application

1067
00:48:10,980 --> 00:48:13,519
depended on identifying a specific syntactic

1068
00:48:13,519 --> 00:48:15,980
domain, like lowering tones only at the end of

1069
00:48:15,980 --> 00:48:18,659
a verb phrase, VP, were mastered significantly

1070
00:48:18,659 --> 00:48:21,320
later. So understanding the syntactic structure

1071
00:48:21,320 --> 00:48:24,119
needed to correctly apply the tone rule is the

1072
00:48:24,119 --> 00:48:27,090
harder part. It appears so. The interface between

1073
00:48:27,090 --> 00:48:30,110
syntax and phonology, especially for complex

1074
00:48:30,110 --> 00:48:32,710
phenomena like tone, seems to be one of the later

1075
00:48:32,710 --> 00:48:35,230
developing aspects of grammar acquisition. Okay,

1076
00:48:35,289 --> 00:48:37,610
we've covered a huge amount of ground, from the

1077
00:48:37,610 --> 00:48:39,869
larynx muscles all the way to infant perception

1078
00:48:39,869 --> 00:48:43,469
and OT constraints. As we wrap up, what are the

1079
00:48:43,469 --> 00:48:45,789
big takeaways here? I think what's really striking,

1080
00:48:45,949 --> 00:48:49,019
synthesizing all this, is seeing tone not just

1081
00:48:49,019 --> 00:48:51,940
as pitch variation, but as this incredibly intricate

1082
00:48:51,940 --> 00:48:55,039
structured system. It's grounded in physical

1083
00:48:55,039 --> 00:48:57,380
anatomy, yes, but it operates through abstract

1084
00:48:57,380 --> 00:49:00,539
phonological tiers. It's governed by universal

1085
00:49:00,539 --> 00:49:03,579
competing pressures like the OCP and faithfulness,

1086
00:49:03,699 --> 00:49:06,219
modeled beautifully by things like ASP and OT.

1087
00:49:06,500 --> 00:49:08,260
And we've seen how different language families

1088
00:49:08,260 --> 00:49:10,219
have exploited this system in really different

1089
00:49:10,219 --> 00:49:12,699
ways, haven't we? Africa focusing on mobility,

1090
00:49:12,960 --> 00:49:15,099
using things like floating tones for downstep.

1091
00:49:15,630 --> 00:49:18,250
Asia, building these massive lexical inventories

1092
00:49:18,250 --> 00:49:20,809
and developing incredibly complex, almost Rube

1093
00:49:20,809 --> 00:49:24,030
Goldberg -esque Sandy chains. Exactly. The regional

1094
00:49:24,030 --> 00:49:26,389
strategies highlight the flexibility within the

1095
00:49:26,389 --> 00:49:28,809
universal toolkit. Which leads us to that final

1096
00:49:28,809 --> 00:49:31,150
provocative thought, maybe. There's this constant

1097
00:49:31,150 --> 00:49:33,699
tension, isn't there? There really is. On one

1098
00:49:33,699 --> 00:49:36,119
hand, we see clear evidence for universal constraints

1099
00:49:36,119 --> 00:49:39,559
rooted in perception and articulation, things

1100
00:49:39,559 --> 00:49:42,639
like the probable five -level limit or the inherent

1101
00:49:42,639 --> 00:49:45,980
difficulty of producing rising tones. These pressures

1102
00:49:45,980 --> 00:49:49,079
should, in theory, push languages towards simpler,

1103
00:49:49,199 --> 00:49:51,500
more natural systems. But then we see languages

1104
00:49:51,500 --> 00:49:54,460
like Shaiman Chinese with its chain shift. Exactly.

1105
00:49:54,500 --> 00:49:57,199
Or Mixtec with its tone -dependent stress and

1106
00:49:57,199 --> 00:50:00,340
quirky floating tone rules. Languages devise

1107
00:50:00,340 --> 00:50:03,070
these wildly complex... sometimes seemingly arbitrary

1108
00:50:03,070 --> 00:50:05,809
systems that fly in the face of phonetic simplicity

1109
00:50:05,809 --> 00:50:08,769
and seem to demand sheer brute force memorization

1110
00:50:08,769 --> 00:50:10,969
from learners. So for you listening, the question

1111
00:50:10,969 --> 00:50:14,090
to ponder is about this balance. How much does

1112
00:50:14,090 --> 00:50:16,449
the universal toolkit, the constraints favoring

1113
00:50:16,449 --> 00:50:19,030
simplicity and naturalness actually shape language?

1114
00:50:19,210 --> 00:50:22,150
And how much is language shaped by, well, historical

1115
00:50:22,150 --> 00:50:24,809
accidents? The specific way constraints got ranked

1116
00:50:24,809 --> 00:50:26,769
in this particular language is history, leading

1117
00:50:26,769 --> 00:50:29,030
to the preservation or even elaboration of a

1118
00:50:29,030 --> 00:50:31,670
really complex marked pattern like that min -chain

1119
00:50:31,670 --> 00:50:34,130
shift. The fact that these constraints are viable,

1120
00:50:34,349 --> 00:50:37,610
that rankings differ, means the world's languages

1121
00:50:37,610 --> 00:50:40,389
showcase this incredible diversity born from

1122
00:50:40,389 --> 00:50:43,210
the collision of physics, perception, grammar,

1123
00:50:43,550 --> 00:50:45,929
and just time. It really makes you wonder what

1124
00:50:45,929 --> 00:50:48,710
other complex, maybe slightly illogical patterns

1125
00:50:48,710 --> 00:50:51,389
exist in language, maybe even in your own language,

1126
00:50:51,550 --> 00:50:53,929
that are only there because, generations ago,

1127
00:50:54,050 --> 00:50:56,269
one specific constraint ranking one out over

1128
00:50:56,269 --> 00:50:58,130
another. It's a fascinating puzzle.