1
00:00:00,000 --> 00:00:05,840
[Music]

2
00:00:05,840 --> 00:00:09,280
Welcome to episode 35 of the Language Neuroscience Podcast.

3
00:00:09,280 --> 00:00:14,480
I'm Stephen Wilson and I'm a neuroscientist at the University of Queensland in Brisbane, Australia.

4
00:00:14,480 --> 00:00:19,440
Just a quick heads up, I have a research position available in my lab right now at the University of Queensland

5
00:00:19,440 --> 00:00:25,760
to work on imaging aspects of our NIH-funded project, Neural correlates of recovery from aphasia after stroke.

6
00:00:25,760 --> 00:00:31,520
This is a level A or B research focused postdoctoral position, which is open to international applicants

7
00:00:31,520 --> 00:00:34,880
and UQ will sponsor a visa for the successful applicant.

8
00:00:34,880 --> 00:00:40,000
If you or anyone you know is interested, please go to langneurosci.org/join to learn more.

9
00:00:40,000 --> 00:00:44,880
Okay, well my guest today will need no introduction for most of our listeners.

10
00:00:44,880 --> 00:00:49,440
I'm really pleased to be joined by Greg Hickok, distinguished Professor of Cognitive Sciences

11
00:00:49,440 --> 00:00:54,160
and Language Science at the University of California, Irvine. Greg is one of the most brilliant and

12
00:00:54,160 --> 00:00:59,120
influential scientists in our field and he has a brand-new book coming out, "Wired for words:

13
00:00:59,120 --> 00:01:04,960
the neural architecture of language", coming out from MIT Press on November 25th, the same day that I'll

14
00:01:04,960 --> 00:01:09,600
release this podcast episode. Today we're going to chat about the book, which lays out Greg's

15
00:01:09,600 --> 00:01:14,080
up-to-date model of language processing in the brain, building on his previous work and of course

16
00:01:14,080 --> 00:01:19,120
150 years of findings in our field. We'll talk about the dual streams, the sensory theory of speech

17
00:01:19,120 --> 00:01:24,800
production, dorsal and ventral precentral speech areas and last but not least our diverging views

18
00:01:24,800 --> 00:01:29,520
on the laterality of the ventral stream. Okay, let's get to it. How's it going?

19
00:01:29,520 --> 00:01:34,080
It's going well. As well as can be trying to do science in the United States these days, but

20
00:01:34,080 --> 00:01:40,800
how are things with you. Pretty good. As you know, I moved to Australia a few years ago

21
00:01:40,800 --> 00:01:46,800
and how about you? Like how's your life going apart from, you know, science challenges?

22
00:01:46,800 --> 00:01:51,840
Life’s pretty good. Life's pretty good. I can't complain. We have a grand kid now.

23
00:01:51,840 --> 00:01:58,080
Oh wow. Yeah. He is already three years old, living in Nashville, oddly enough.

24
00:01:58,080 --> 00:02:03,360
Oh, okay. Older daughter lives in Nashville. What's she doing there?

25
00:02:03,360 --> 00:02:10,800
She's, they were in Arizona and didn't want to be in Arizona and didn't want to come to California

26
00:02:11,360 --> 00:02:16,880
and settled on Nashville as a place. Okay. Is it affordable and interesting in some ways, so…

27
00:02:16,880 --> 00:02:22,880
It is. Well, it's not affordable anymore, but you know, so your daughter has like really replicated

28
00:02:22,880 --> 00:02:28,960
my life of moving from California to Arizona to Nashville. (Laughter) Yeah, exactly. Yeah, clearly that was

29
00:02:28,960 --> 00:02:35,680
yeah. Maybe she'll hit Australia next. It'll be good for her. Not so good, but maybe not for you.

30
00:02:38,000 --> 00:02:43,280
Yeah. So, are you still getting to the beach a lot and taking advantage of Southern California?

31
00:02:43,280 --> 00:02:50,560
Oh, yeah. Yep. I head down to the beach a couple times a week, not surfing much anymore,

32
00:02:50,560 --> 00:02:58,240
but paddling. I have like a racing style canoe, like the surf skis they call them, that are just

33
00:02:58,240 --> 00:03:04,240
kind of like an outrigger without the outrigger. So, I go paddle that around for a few hours

34
00:03:04,240 --> 00:03:09,760
a couple times a week. Okay. So, that's more your speed than surfing these days? Well, yeah, it's just

35
00:03:09,760 --> 00:03:15,040
easier to get out there and you know, it's gotten crowded in the water and sharky, you know, it's gotten,

36
00:03:15,040 --> 00:03:21,200
it's, you see sharks out there now. Oh, wow! More like Australia. So, yeah, great whites. So,

37
00:03:21,200 --> 00:03:25,840
oh, goodness. Yeah, I'm good with, with paddling and canoeing around here and out.

38
00:03:25,840 --> 00:03:32,240
Kayak around. Okay. Yeah, there's been quite a few shark attacks in Australia this year.

39
00:03:32,960 --> 00:03:39,680
Pretty, pretty shocking ones sometimes. But it doesn't stop my family from playing in the water.

40
00:03:39,680 --> 00:03:49,280
Right. Yeah. Yeah, exactly. Okay. So, you've got a book that's coming out. We're going to chat about

41
00:03:49,280 --> 00:03:53,840
that. When is the, do you know when the book is coming out? Like when's it going to be officially released?

42
00:03:53,840 --> 00:03:59,760
Yes, it’s coming out I think the 24th or something. Okay. A couple weeks from now.

43
00:03:59,760 --> 00:04:04,720
Okay. So, just for context, we're talking on the 13th for me, 12th for you.

44
00:04:04,720 --> 00:04:08,720
And so, yeah, it usually takes me at least a week to edit anyway. So, we'll

45
00:04:08,720 --> 00:04:13,120
release our conversation on the same day that your book gets released.

46
00:04:13,120 --> 00:04:18,960
That's awesome. And yeah, it was really fortuitous because like I emailed you to see if you wanted

47
00:04:18,960 --> 00:04:24,960
to chat and not knowing that you had the book coming out and then learned that you did and

48
00:04:24,960 --> 00:04:30,480
that was perfect because I've got a chance to, you know, read your current opinions about all these

49
00:04:30,480 --> 00:04:37,120
things that you've been working on for 25, 30 years. And then it gives us kind of, you know, a big,

50
00:04:37,120 --> 00:04:42,880
big picture thing to talk about. So, yeah, thanks for making the time to talk with me.

51
00:04:42,880 --> 00:04:48,560
Yeah, of course. I'm excited to do it. I think you're the first person I've talked to about the book

52
00:04:48,560 --> 00:04:57,840
besides a few couple students. Oh, that's cool. Yeah. Media. The first and only podcast about the

53
00:04:57,840 --> 00:05:06,800
scientific study of language in the brain. So, what is your book? Is it a textbook, a memoir,

54
00:05:06,800 --> 00:05:13,600
or something in between? I think it's definitely not a textbook, although it's, so I wrote it,

55
00:05:13,600 --> 00:05:18,480
I wrote it with my students in mind. I've been teaching this course, Language in the Brain at UCI,

56
00:05:18,480 --> 00:05:26,160
since 1997. And there's no textbook; there wasn't what I thought of writing this book and I

57
00:05:26,160 --> 00:05:33,120
decided I would write a textbook myself and pitched it to MIT Press back in probably 1999 and

58
00:05:33,120 --> 00:05:41,040
it got approved and I got a contract around the year 2000 to do it. But then that was around the

59
00:05:41,040 --> 00:05:45,120
time when David Poeppel and I were developing the dual street model and I'm like, oh my god,

60
00:05:45,120 --> 00:05:49,600
I had to figure out how this stuff is working before I can write a textbook on brain and language.

61
00:05:49,600 --> 00:05:54,800
And then time got away, and I kept working on other things and

62
00:05:54,800 --> 00:06:03,680
wrote a book in between but then decided I owed MIT Press this book. So, I got back to it in 2016 or so.

63
00:06:03,680 --> 00:06:09,360
And by then David Kemmerer had his book out textbook, which is quite good and there was no point

64
00:06:09,360 --> 00:06:15,200
of writing another textbook. So, I decided to make it a kind of, that's a pretty high paise, to David

65
00:06:15,200 --> 00:06:19,440
Kemmerer, by the way. Like, he'd already written one, so there's no point writing another one, but yeah.

66
00:06:19,440 --> 00:06:24,720
Yeah, I thought it was quite good. David and I joke that his description of the dual stream model

67
00:06:24,720 --> 00:06:31,040
was way better than ours. (Laughter) We thought it was pretty, you know, it was pretty good. So yeah, no one

68
00:06:31,040 --> 00:06:36,960
like his textbook too. Yeah, I asked MIT Press if I could editor there, if I could just write a

69
00:06:36,960 --> 00:06:46,480
monograph kind of my view of the world of brain and language and I wrote it basically as the

70
00:06:46,480 --> 00:06:53,040
following the content of my course and tried to make it accessible to undergraduates at least

71
00:06:53,040 --> 00:06:58,320
with some guidance. So that was kind of my target for this book. Yeah, so it's kind of in between.

72
00:06:58,320 --> 00:07:03,600
Yeah, and so having taken a look at it and I can share with our listeners that,

73
00:07:03,600 --> 00:07:12,720
okay, first of all, you should definitely read it. I loved it. It's for many reasons. One of which is

74
00:07:12,720 --> 00:07:19,760
just it's written in such a accessible conversational way, right? So, like I so I know you and I've been

75
00:07:19,760 --> 00:07:26,240
chatting with you for decades, believe it or not. But I really when I read it, I just hear your voice,

76
00:07:26,240 --> 00:07:31,120
you know, like it's just like this is Greg talking and it doesn't, it doesn't read stuffy at all.

77
00:07:31,120 --> 00:07:35,040
It just kind of reads like, well, there's what I thought and then I thought I might be wrong. So, then we

78
00:07:35,040 --> 00:07:40,800
tried this and it turned out I wasn't wrong. I was definitely right all along. (Laughter) So, you know what I mean?

79
00:07:40,800 --> 00:07:44,480
Yeah, yeah. Did you do that deliberately or is that just the only way you can write?

80
00:07:44,480 --> 00:07:51,360
It's kind of my style. But yeah, I did want to make it accessible and interesting. I don't

81
00:07:51,360 --> 00:07:56,240
stuffy textbooks are just kind of boring to read and scientific papers are kind of boring to read.

82
00:07:57,520 --> 00:08:02,080
And I just wanted something where I could go through this stuff without, without pulling punches on

83
00:08:02,080 --> 00:08:07,840
the details, but trying to make it accessible to people who were interested, really interested in

84
00:08:07,840 --> 00:08:13,040
understanding bits and pieces about this system worked. And it's really not just for it. It's not

85
00:08:13,040 --> 00:08:17,840
just for undergrads by any means, right? I mean, I think that there's like, you know, I think that

86
00:08:17,840 --> 00:08:24,160
all of our colleagues in the field could take a lot from this book. Yeah, that's right. I wrote it

87
00:08:24,160 --> 00:08:31,200
accessible to undergraduates, but I when I was thinking I was writing these arguments and

88
00:08:31,200 --> 00:08:37,920
referencing everything so that it would be a perfectly legitimate argument for the practicing

89
00:08:37,920 --> 00:08:44,320
scientists for sure. Yeah, cool. So, I don't know if you've ever listened to my podcast, but I always

90
00:08:44,320 --> 00:08:48,720
kind of like to talk to people about how they got into the field, how they developed their interests.

91
00:08:48,720 --> 00:08:52,800
And you sort of tell us a bit in the start of the book and I'll read a quote here. It says,

92
00:08:53,360 --> 00:08:57,200
"This is a book about the biology of language, my primary area of research.

93
00:08:57,200 --> 00:09:00,720
Oddly enough, I never much liked the topic during my college years."

94
00:09:00,720 --> 00:09:07,760
And then you go on to explain how you got interested in the neuroscience of language. So can you just

95
00:09:07,760 --> 00:09:15,520
share that back story with us? Yeah, sure. So, I was interested in mind and brain as an undergraduate,

96
00:09:15,520 --> 00:09:22,720
and anything mind and brain fascinated me except the language stuff. I was in psychology classes,

97
00:09:22,720 --> 00:09:28,480
because you take classes and learn about language. And a lot of it is just like, "Oh, there's these

98
00:09:28,480 --> 00:09:32,400
nouns and verbs, and then there's phrase structure, and there's this and that and it's just like,

99
00:09:32,400 --> 00:09:40,000
yeah, it never excited me. It was nothing I was interested in. So, yeah, it didn't, it bored me,

100
00:09:40,000 --> 00:09:45,200
and like I said in the book, I remember picking up a copy of a neuropsychology book and

101
00:09:45,200 --> 00:09:50,000
realizing that if I wanted to do neuropsychology someday, I'd probably have to learn about these

102
00:09:50,000 --> 00:09:55,760
aphsias and stuff. And I was like, all right, fine, I'll learn about it, whatever. But yeah, but then I

103
00:09:55,760 --> 00:10:03,280
happened to get into grad school at Brandeis University where the advisor that I would be, a, work with

104
00:10:03,280 --> 00:10:08,400
doing neuropsychology was Edgar Zurif. He was a famous aphasiologist, and I was like, Okay, I guess

105
00:10:08,400 --> 00:10:15,600
I'm learning language. But how did you get paired up with Zurif if you weren't interested in language?

106
00:10:17,440 --> 00:10:23,120
That's an accident, actually. So, I was applying to grad school. I applied to MIT, Johns Hopkins,

107
00:10:23,120 --> 00:10:29,840
I wanted to work with the neuropsychologist in those kinds of places. My undergraduate advisor

108
00:10:29,840 --> 00:10:35,920
was someone named Mary Louise Cain, who was an aphasiologist, and I was doing face perception

109
00:10:35,920 --> 00:10:41,680
research with her because that was interesting to me. And I was applying to programs, and she said,

110
00:10:41,680 --> 00:10:46,560
Hey, you should apply to Brandeis. That's a good program, and Edgar Zurif is there. And I was like,

111
00:10:46,560 --> 00:10:51,840
Well, I don't really want to do language, but okay. And that's the program I got into. I didn't get

112
00:10:51,840 --> 00:10:57,280
in at any other place. And so that's where I went. And so, I started learning language.

113
00:10:57,280 --> 00:11:04,080
That's like just an interesting little side note there. Isn't it? For those listeners who are struggling

114
00:11:04,080 --> 00:11:07,920
to get into grad school and wondering what grad school they're going to get into and sending

115
00:11:07,920 --> 00:11:14,960
all these applications. You only got into one grad school and then go on to become you. (Laughter)

116
00:11:14,960 --> 00:11:21,840
Yeah, it was my backup plan, and I got in and said, Okay, I'm going to do this, I guess.

117
00:11:21,840 --> 00:11:27,920
That's great. That's a great story. So, I mean, so did Edgar you interested in it?

118
00:11:27,920 --> 00:11:34,000
You know who got me interested in it? I tried to read some of his papers before arriving in

119
00:11:34,000 --> 00:11:39,920
Massachusetts in Boston. And it was just this still wasn't that interesting to me. I was reading his

120
00:11:39,920 --> 00:11:46,320
papers and, alright, whatever. But I took a course on graduate syntax by Jane Grimshaw.

121
00:11:46,320 --> 00:11:56,560
And she's a well-known linguist doing syntax. And I just jumped in, and she taught the course and

122
00:11:56,560 --> 00:12:00,960
I remember the very first assignment, homework assignment I got where we had to analyze some

123
00:12:00,960 --> 00:12:07,280
sentence structure. And I remember going to one of my fellow grad students and saying,

124
00:12:07,280 --> 00:12:11,920
I don't even have any idea what format an answer is supposed to look like. It was a phrase

125
00:12:11,920 --> 00:12:18,640
structure kind of set of rules that they were looking for that she was looking for. And that was my

126
00:12:18,640 --> 00:12:24,400
beginning. But what I learned from Jane was really important things. That language is this kind

127
00:12:24,400 --> 00:12:29,760
of internal system that generates these things. It's part of how the mind works and how language

128
00:12:29,760 --> 00:12:36,640
gets learned and processed. And I realized that it was really interesting. That it wasn't what I

129
00:12:36,640 --> 00:12:42,400
thought it was. And so, I got interested. And then kind of went into hardcore, not syntax syntax,

130
00:12:42,400 --> 00:12:48,320
but I did straight up psycholinguistics. And when I finally did start diving into aphasiology,

131
00:12:48,320 --> 00:12:56,160
it was grounded in serious, Chomskyan kind of theoretical syntax, if you read my early papers.

132
00:12:56,160 --> 00:13:03,520
Yeah, I have done. Sorry, my dog's just decided to start barking at nothing probably.

133
00:13:03,520 --> 00:13:12,720
Molly! I can edit that out. I can't even hear it. Yeah, Zoom cancels these things.

134
00:13:12,720 --> 00:13:18,480
Okay, so yeah, no, I've read those papers of yours from the late 90s. It doesn't read like later

135
00:13:18,480 --> 00:13:27,760
Greg, but like it's like formative, right? Yeah, yeah, I was another interesting story. So, I took,

136
00:13:27,760 --> 00:13:35,200
as a third year graduate student at Brandeis, I took a course by Steve Pinker and Ellen Prince,

137
00:13:35,200 --> 00:13:41,520
who were then all about the past tense and, you know, connectionist and critiques and all that sort of

138
00:13:41,520 --> 00:13:50,400
thing. And as my project, my paper project for that course, I did a paper proposing an experiment

139
00:13:50,400 --> 00:13:58,960
in aphasia studying past tense production. And Steve liked it and eventually invited me to

140
00:13:58,960 --> 00:14:04,560
become a postdoc at MIT with him, the new McDonnell-Pew Center for Cognitive Neuroscience,

141
00:14:04,560 --> 00:14:10,240
which I went and did. But I wasn't terribly interested in doing aphasiology,

142
00:14:10,240 --> 00:14:15,040
ended up working with Ted Gibson on some straight up psycholinguistic stuff. And at the end of my

143
00:14:15,040 --> 00:14:21,440
first year, they were going to kick me out of MIT because I wasn't doing neuroscience, but I was

144
00:14:21,440 --> 00:14:25,520
just doing straight up kind of cognitive stuff. Why didn't they kick Ted out then? (Laughter)

145
00:14:25,520 --> 00:14:31,840
Well, he, yeah, no, he was a graduate student, no, he was a postdoc. I don't know, he probably should

146
00:14:31,840 --> 00:14:38,080
have. (Laughter) He might have been in under a different program or something, I don't remember. But anyway,

147
00:14:38,080 --> 00:14:44,320
so, I decided that I would start doing syntax and brain and that's kind of the, they told me that I

148
00:14:44,320 --> 00:14:49,840
could stay another year if I just started doing that. So that's how it started. Yeah, I didn't know

149
00:14:49,840 --> 00:14:54,960
that. I didn't know that back story. That's interesting. So then, so that's how you kind of got into it.

150
00:14:54,960 --> 00:15:01,840
Now, in the book you talk about, in the preface, you say, you have a tendency, I wouldn't

151
00:15:01,840 --> 00:15:08,320
quote again, like, you have a tendency to frame the issues in historical context. And I really like that

152
00:15:08,320 --> 00:15:16,080
approach. That's, that's the way that I think too. But as you note, I quote, some readers don't necessarily

153
00:15:16,080 --> 00:15:20,400
want this kind of framing, preferring instead to just get the facts of the modern understanding without

154
00:15:20,400 --> 00:15:26,480
mulling over the old ideas and results. So why do you think it's good, why do you think it's good

155
00:15:26,480 --> 00:15:33,360
to think this way? And why do you think this way? I naturally like history. So, I'll go back and see,

156
00:15:33,360 --> 00:15:38,720
you know, what happened back in the day. That's just fun for me. But I found that it's useful because

157
00:15:38,720 --> 00:15:45,520
it shows you things about where these ideas came from and what kinds of ideas got rejected,

158
00:15:45,520 --> 00:15:50,560
why, and sometimes for the wrong reasons. So sometimes what happens in a field is,

159
00:15:50,560 --> 00:15:56,080
is someone comes along and says, Oh, no, that's a bad idea. And here's why and everyone buys it. But

160
00:15:56,080 --> 00:16:03,600
the argument is bogus. But then the idea is gone, and everyone's forgotten it. So, going back and seeing

161
00:16:03,600 --> 00:16:09,760
what the ideas were, where they came from originally, why they were rejected or accepted. And then,

162
00:16:09,760 --> 00:16:14,720
you know, moving forward, it gives you a broader perspective on the various, the range of ideas and

163
00:16:14,720 --> 00:16:20,880
arguments. And I found plenty of flaws in old arguments as I try to detail in the book

164
00:16:23,120 --> 00:16:29,040
that helped me understand why we have the biases that we do today and how we can move forward

165
00:16:29,040 --> 00:16:36,640
most productively. So, I'm happy to tell. Yeah, I appreciate that framing of everything in your

166
00:16:36,640 --> 00:16:44,800
in your book and in your work in general. And so again, at the start of the book, you talk about

167
00:16:44,800 --> 00:16:49,920
how the human brain is specialized for language, which is something that we would obviously agree on

168
00:16:49,920 --> 00:16:54,720
and the most of our listeners would probably agree on. And you raised this interesting question about

169
00:16:54,720 --> 00:17:01,120
whether it evolved from scratch or did it evolve by tinkering with what was lying around. So, can

170
00:17:01,120 --> 00:17:04,960
you kind of share like, what's your perspective on that and what ramifications does it have for

171
00:17:04,960 --> 00:17:12,400
understanding the neurobiology of language? Yeah, so it's always been difficult to be a student

172
00:17:12,400 --> 00:17:16,800
of the neurobiology of language. Whereas if you're a setting vision, we can go and study cats or

173
00:17:16,800 --> 00:17:23,440
macaques or we have all this information from other homologous visual systems where we can kind of

174
00:17:23,440 --> 00:17:28,160
jumpstart the process of understanding how things are working in humans. I grew up with the

175
00:17:28,160 --> 00:17:36,240
the idea, the belief that there was no homologue. There was no animal model for language.

176
00:17:36,240 --> 00:17:42,480
You know, obviously, because we're the only speaking animal, at least at the level of sophistication

177
00:17:42,480 --> 00:17:50,080
that we have in the world. So how do we make progress? And then if that's true for the only speaking

178
00:17:50,080 --> 00:17:55,600
animal or linguistic animal, how did we get it? How did it start? And that's been, of course, a fundamental

179
00:17:55,600 --> 00:18:04,640
question that people had wondered about for a long time. So, it's just a fundamental question about

180
00:18:04,640 --> 00:18:10,960
how language evolved. And the problem is that evolution doesn't really work like that. It doesn't

181
00:18:10,960 --> 00:18:16,560
work by inventing brand new things. It tinkers with existing systems, modifies them, and it can

182
00:18:16,560 --> 00:18:23,680
modify them through descent quite substantially so that they become quite unique. But they have to

183
00:18:23,680 --> 00:18:27,840
come from somewhere. So where did language come from? And people that propose various ideas. But

184
00:18:27,840 --> 00:18:35,840
yeah, that's kind of the, the basic problem with that issue. And why I think the approach that I've

185
00:18:35,840 --> 00:18:43,120
taken kind of helps solve that in some ways. Yeah. So, you basically, you're perspective, and which I

186
00:18:43,120 --> 00:18:50,160
share is that it evolved by tinkering. And that ends up having, then as you kind of explore all the

187
00:18:50,160 --> 00:18:55,360
different aspects of the language network, you always kind of relate them to other principles of

188
00:18:55,360 --> 00:19:00,960
neuroscience or principles of brain organization in different modalities and systems. Yeah.

189
00:19:02,480 --> 00:19:09,040
Yeah, for sure. I think this was a natural thing for me to do. I didn't consciously

190
00:19:09,040 --> 00:19:17,360
start doing this. But back in the 90s, I was reading Milner and Goodale’s book on the two visual

191
00:19:17,360 --> 00:19:25,040
streams, the visual brain in action, I think it's called. And I saw a lot of parallels to language.

192
00:19:25,040 --> 00:19:30,880
And so that's kind of drew me to start thinking about language in those terms. And it helped me

193
00:19:30,880 --> 00:19:38,320
organize how I thought about very a wide range of data in aphasiology and in functional imaging.

194
00:19:38,320 --> 00:19:43,520
And ultimately led to the dual stream model that David and I proposed in the 2000s.

195
00:19:43,520 --> 00:19:49,360
And so, I realized that that kind of approach, thinking in terms of evolutionary homologies,

196
00:19:49,360 --> 00:19:55,840
was helping me think of things that I wouldn't have thought of otherwise and to develop hypotheses to

197
00:19:55,840 --> 00:20:04,560
test. And they turned out to be reasonably accurate in my view. So yeah. I think it's a really helpful

198
00:20:04,560 --> 00:20:09,840
approach. Yeah, cool. So, let's kind of start talking about the model that you develop in this book.

199
00:20:09,840 --> 00:20:14,960
And it's got several different components to it. But like you just mentioned the dual stream model.

200
00:20:14,960 --> 00:20:19,840
And obviously that's what you're very well known for. And it's also a prominent part of this book.

201
00:20:21,440 --> 00:20:28,560
You know, your 2007 paper has almost 7000 citations and the earlier papers have many thousands as well.

202
00:20:28,560 --> 00:20:35,680
And this kind of like pervades, I'd say the first half of the book. So can you talk about,

203
00:20:35,680 --> 00:20:41,120
you know, the dual stream concept in the visual domain just for, you know, briefly for those that

204
00:20:41,120 --> 00:20:48,800
are familiar with it and then kind of talk about how you develop those concepts in the language domain.

205
00:20:49,680 --> 00:20:56,320
Yeah, sure. So, in the visual domain, it's conceptually the argument is pretty simple. So, as you're looking at

206
00:20:56,320 --> 00:21:01,280
Stephen's image in the video here, you see a microphone sitting there in front of him.

207
00:21:01,280 --> 00:21:07,200
And there's two things you can do with that information. You know, conceptually, one is that you can

208
00:21:07,200 --> 00:21:12,880
look at it and understand what it is and map it to some semantic representation and understand that it is a

209
00:21:12,880 --> 00:21:18,800
a recording amplification device and it's doing something. It's meaningful to you. On the other

210
00:21:18,800 --> 00:21:24,240
hand, if you are in the room with Stephen, you could reach out and grab it. Now, those two tasks are

211
00:21:24,240 --> 00:21:29,920
fundamentally different. It's going to be a microphone, whether it is in its place now, whether it was

212
00:21:29,920 --> 00:21:36,080
above Stephen's head, lying on the table, upside down, it's still a microphone, no matter its position

213
00:21:36,080 --> 00:21:41,520
in space or orientation, and maybe there's a little mini microphone. That's, you know, same semantic

214
00:21:41,520 --> 00:21:48,080
category, but different size and shape. So, it doesn't change the semantic content, but what it does do is

215
00:21:48,080 --> 00:21:53,040
change the way you might interact with that if you're going to reach and pick it up. So, all those features

216
00:21:53,040 --> 00:21:58,560
that don't matter for categorizing semantically, suddenly matter for everything when you're trying to

217
00:21:58,560 --> 00:22:05,840
reach and grasp it. So, the idea there is that the brain perceives information and processes it

218
00:22:05,840 --> 00:22:11,440
along two different streams, one for concept, conceptual understanding. That's the ventral stream

219
00:22:11,440 --> 00:22:16,400
and one for motorically interacting with that object. So that's ‘the what’ and ‘the how’ stream

220
00:22:16,880 --> 00:22:22,720
respectively. So that's the visual domain and arguments were laid out nicely by Milner and Goodale.

221
00:22:22,720 --> 00:22:30,160
I encourage folks to read it. And it occurred to me that something very similar is happening in

222
00:22:30,160 --> 00:22:34,640
in speech and language. So, as you're listening to my words, there's two different things you can do

223
00:22:34,640 --> 00:22:43,600
with them. You can hear them for and understand them, or you can take those perceptual signals, those

224
00:22:43,600 --> 00:22:49,200
phonemic sequences and map them onto your own motor system so that you can say the same words. And

225
00:22:49,200 --> 00:22:55,600
obviously, we do that in development as part of learning because you hear speech words and speech

226
00:22:55,600 --> 00:23:00,480
sounds in the environment and you figure out not only how to make sense of them in terms of their

227
00:23:00,480 --> 00:23:05,360
meaning, but you figure out how to be able to reproduce those words in those phonemic patterns

228
00:23:05,360 --> 00:23:12,960
yourself. So those are two separate kind of mappings, which we hypothesized were a ventral and

229
00:23:12,960 --> 00:23:21,440
a dorsal stream system. And that was kind of the foundation for thinking about speech processing

230
00:23:21,440 --> 00:23:27,680
in terms of these two streams. Yeah, cool. And so, you know, you talk about that historical

231
00:23:27,680 --> 00:23:33,040
perspective that you bring to everything, and you know in your papers from the first decade of

232
00:23:33,040 --> 00:23:38,800
century as well as in this book, the relationship between the dual stream concept and

233
00:23:38,800 --> 00:23:43,200
Wernike-Lichtheim model. So can you talk a bit about like, you know, to what extent were those old

234
00:23:43,200 --> 00:23:49,520
guys barking up the right tree and what and where did they kind of not quite have it right?

235
00:23:49,520 --> 00:23:57,120
Yeah, so Wernike sometimes doesn't get respect in the modern world. We all know that it's not

236
00:23:57,120 --> 00:24:03,120
right in all details and there's a tendency when you know models aren't working very well for this

237
00:24:03,120 --> 00:24:09,200
or that you toss them out. They don't do anything for syntax, or you know any of the more complicated

238
00:24:09,200 --> 00:24:15,360
linguistic things that we think about these days. But they were onto something. That Wenike’s

239
00:24:15,360 --> 00:24:20,880
model was the first dual stream model that predates the visual. We always think of the Hickok-Poeppel

240
00:24:20,880 --> 00:24:27,520
dual stream model is derivative of the, I always thought this derivative of the Milner-Goodale

241
00:24:27,520 --> 00:24:32,320
two visual streams going back further, Ungeleider and Mishkin model with the ‘What, Where’ system.

242
00:24:33,120 --> 00:24:38,800
But Wernike proposed essentially a dual stream model where you had information coming into

243
00:24:38,800 --> 00:24:43,760
the auditory cortex. It was sent in two different directions. One was to the motor system, one was to

244
00:24:43,760 --> 00:24:49,600
the conceptual system. That is essentially the dual stream model. And in fact, our dual stream model

245
00:24:49,600 --> 00:24:55,840
was merely an elaboration and reframing and updating of Wernike’s original model. So, I don't

246
00:24:55,840 --> 00:25:00,800
take credit for that at all. I think it was already there in the literature. And he even talked about

247
00:25:00,800 --> 00:25:08,560
important concepts like what we today call feedback control in motor planning. He talked about it in

248
00:25:08,560 --> 00:25:14,720
terms of a corrective function of the auditory system on motor cortex. So, it was quite advanced at the

249
00:25:14,720 --> 00:25:22,320
time. Absolutely. Yeah. I like how you always draw those connections throughout the book to the

250
00:25:22,320 --> 00:25:29,280
history of all the ideas that you talk about. So, I think that probably most people would,

251
00:25:30,000 --> 00:25:34,080
I'd like to talk more about the ventral stream. And that's kind of the question that I raised

252
00:25:34,080 --> 00:25:41,040
that I wanted to talk to you about, especially laterality. But maybe saving that for later in

253
00:25:41,040 --> 00:25:48,400
our discussion. But moving forward with the dorsal stream that perhaps is sometimes a less of a focus

254
00:25:48,400 --> 00:25:53,600
of other people reading your work, but I think might be very much a focus of your own interest,

255
00:25:53,600 --> 00:25:59,280
especially in this book. I think you really develop your dorsal stream ideas more. So, can you

256
00:25:59,920 --> 00:26:04,320
let's, can we talk about what goes on in the dorsal stream and you kind of describe it as the

257
00:26:04,320 --> 00:26:08,320
sensory theory of speech production. So, what's that all about?

258
00:26:08,320 --> 00:26:14,480
Yeah. So, one important thing just as a prerequisite before we start talking about this is the

259
00:26:14,480 --> 00:26:20,960
dual stream model as David and I developed it and as later, Rauschecker and Scott developed it

260
00:26:20,960 --> 00:26:28,720
in their version of this, it only makes sense from the perspective of auditory perception. So the

261
00:26:28,720 --> 00:26:34,480
dual stream model is a very perceptually grounded model. And I think people misunderstood it and took it

262
00:26:34,480 --> 00:26:40,640
as a model of language processing in general. Like some people I remember would say, oh, the

263
00:26:40,640 --> 00:26:45,040
dorsal stream is involved in phonological processing. So, the way that we comprehend words

264
00:26:45,040 --> 00:26:50,080
is we take phonological information, process it in the dorsal stream, and then you come back into

265
00:26:50,080 --> 00:26:55,760
the ventral stream for comprehension. But that's not at all what we were saying. So, it's important to

266
00:26:55,760 --> 00:27:02,640
think about the dual stream model as making sense from the perspective of the auditory system.

267
00:27:02,640 --> 00:27:06,480
From the auditory system, speech information comes in and it can go in two different ways.

268
00:27:06,480 --> 00:27:13,120
If we're just talking about producing speech, naming pictures, natural volitional production,

269
00:27:13,120 --> 00:27:16,880
dual stream model doesn't make much sense because you're going from concepts wherever they're

270
00:27:16,880 --> 00:27:22,880
represented and we have ideas that I laid out in the book to kind of morphosyntactic or

271
00:27:22,880 --> 00:27:29,120
lemma level things to phonological level things and then on out to the motor system in for speech

272
00:27:29,120 --> 00:27:37,040
motor planning and such. And that's just one stream from concept to lemmas to phonological output.

273
00:27:37,040 --> 00:27:41,600
And there's no dual stream involved at all. It involves in fact both of the systems that we

274
00:27:41,600 --> 00:27:46,160
talk about in terms of the dual stream, the ventral and the dorsal part. So that's the first thing.

275
00:27:46,720 --> 00:27:54,080
So, what's the point of the dorsal stream? So, the idea there is that it was mostly developed to

276
00:27:54,080 --> 00:27:59,200
kind of explain the architecture of the phonological system for speech production.

277
00:27:59,200 --> 00:28:06,000
And the basic idea is derived from another field motor control, that motor control theory,

278
00:28:06,000 --> 00:28:12,080
which is all about hitting sensory targets with your actions. So again, let's think about

279
00:28:12,080 --> 00:28:19,280
reaching and grasping. So, if you're going to reach for Stephen’s microphone, you are taking a sensory

280
00:28:19,280 --> 00:28:25,120
information about its location and its orientation and its size and you're using that to guide some action

281
00:28:25,120 --> 00:28:31,520
towards it. So, it's if you didn't know that sensory information, you could never plan the sensory

282
00:28:31,520 --> 00:28:40,240
movements or the motor actions to do the coding. So, movement planning is grounded in sensory systems

283
00:28:40,240 --> 00:28:46,480
in that sense. So, the idea is that linguistic processing is grounded in a similar way. So when

284
00:28:46,480 --> 00:28:53,520
we're producing a word for a microphone, say, even though it's not physically out in the environment,

285
00:28:53,520 --> 00:28:58,560
there is a sensory or an auditory related code for that sound pattern that's stored in your brain

286
00:28:58,560 --> 00:29:04,320
because you've heard that word over again in auditory related cortex. And so that is the target

287
00:29:04,320 --> 00:29:09,760
that your the sensory related target that your phonological planning system is aiming to hit.

288
00:29:09,760 --> 00:29:17,120
So to speak. And so it relies on this posterior auditory related target that is then translated into

289
00:29:17,120 --> 00:29:22,640
motor phonological plans and then executed. So that's kind of what the function of the dorsal

290
00:29:22,640 --> 00:29:29,840
stream is, is to integrate posterior targets that are sensory related in an abstract way

291
00:29:29,840 --> 00:29:38,000
to motor plans for hitting those targets. Yeah. Okay, that's so fundamental. And you know,

292
00:29:38,000 --> 00:29:42,960
you give this really nice example in the book that I think drives home why this is the right way to

293
00:29:42,960 --> 00:29:48,080
think about things. And it involves putting a pencil in your mouth. And so, like our listeners are

294
00:29:48,080 --> 00:29:52,000
not going to see this because it's an audio podcast, but I have a pencil on going to put it in my mouth.

295
00:29:52,000 --> 00:29:57,840
And your point is that you put a pencil in your mouth that you can still talk. And so I'm going

296
00:29:57,840 --> 00:30:02,160
to try that right now. I'm going to put the pencil in my mouth. Okay, now holding a pencil

297
00:30:02,160 --> 00:30:08,720
return why taste and I can still quite effectively produce a sentence that should be intelligible.

298
00:30:08,720 --> 00:30:14,880
And that shows and I had to make some very dramatic motor accommodations to achieve that which I

299
00:30:14,880 --> 00:30:21,200
was completely, which was completely subconscious obviously. But I think it's a powerful demonstration

300
00:30:21,200 --> 00:30:28,000
of this concept that like in speech production we are trying to reach sensory goals and that we've

301
00:30:28,000 --> 00:30:34,400
got this system that's very flexibly able to do that. Yeah. Yeah, that's right. That's a good example.

302
00:30:34,400 --> 00:30:41,840
We it's not just that we are controlling the position and trajectory of articulators. We are

303
00:30:41,840 --> 00:30:46,960
aiming for an auditory target which is a point that Frank Guenther has made repeatedly and done a

304
00:30:46,960 --> 00:30:52,480
great job demonstrating that. And you see that in lots of other paradigms too like ultra-auditory

305
00:30:52,480 --> 00:30:57,600
feedback (UAF), where we will quickly accommodate to perturbations and what we think we hear

306
00:30:57,600 --> 00:31:04,000
or say or what we are hearing us ourselves saying artificially. So yeah, lots of evidence for that.

307
00:31:04,000 --> 00:31:10,000
But the pencil test is the easiest demonstration. Yeah, it makes the point very strongly. So how did

308
00:31:10,000 --> 00:31:18,080
you get interested in how did you come to work on this particular question of like how these auditory

309
00:31:18,080 --> 00:31:25,440
motor transformations in speech production? Yeah, so we had proposed the dual stream model and the

310
00:31:25,440 --> 00:31:30,960
dorsal stream involved three main components. There was like a phonological target system which is

311
00:31:30,960 --> 00:31:37,360
basically we think of as the superior,  posterior superior temporal sulcus which stores the sound related

312
00:31:37,360 --> 00:31:43,120
phonological representations of words. Then there's motor phonological plans that is the motor plans

313
00:31:43,120 --> 00:31:48,240
whatever format they are in aimed at hitting those targets and a translation system in between.

314
00:31:48,240 --> 00:31:53,200
And the reason why I ended up focusing on the dorsal stream so much is because that was the one

315
00:31:53,200 --> 00:31:59,440
part of the model that we didn't have direct evidence for. We proposed an in-between area that

316
00:31:59,440 --> 00:32:05,200
was doing this transformation and that idea came directly out of monkey research on visual motor

317
00:32:05,200 --> 00:32:11,760
grasping where they had identified regions in the intraparietal sulcus (IPS), that were involved in

318
00:32:11,760 --> 00:32:18,720
transforming visual inputs into motor plans for controlling eye movements or grasping actions.

319
00:32:18,720 --> 00:32:24,800
So, we thought well if speech works the same way there's got to be a translation system in between

320
00:32:24,800 --> 00:32:30,640
an integration area that's doing this kind of coordinate transformation as they call it in the

321
00:32:30,640 --> 00:32:39,200
visual world. And so we went looking for it in fMRI. So, we spent a good chunk of the 2000s identifying

322
00:32:39,200 --> 00:32:45,200
this circuit. We called that translation system area SPT for Sylvian, parietal, temporal. Although the

323
00:32:45,200 --> 00:32:50,720
lab we just called it informally in the lab the ‘spot’ and that's how we reverse engineered the

324
00:32:50,720 --> 00:33:02,000
acronym SPT. (Laughter) So, we had identified this and done what I think is a pretty decent job in arguing that

325
00:33:02,000 --> 00:33:07,200
it was doing it was doing in some sensory motor auditory motor function. But then people would

326
00:33:07,200 --> 00:33:11,920
always say well what do you mean by transformation? What do you mean by integration? And I would always

327
00:33:11,920 --> 00:33:17,760
say I don't know yet. I'm just working on establishing the area first and then I'll figure out what is

328
00:33:17,760 --> 00:33:23,680
doing computationally. So, by the end of the 2000s I decided to turn my attention to what it's doing

329
00:33:23,680 --> 00:33:29,680
computationally. And I had two places to look. One was psycholinguistics where you had detailed models

330
00:33:29,680 --> 00:33:36,960
from Pim Levelt and Gary Dell and other people about how we produce speech. And then you had the motor

331
00:33:36,960 --> 00:33:42,960
control people like Frank Guenther and many others who were doing experiments on

332
00:33:42,960 --> 00:33:46,880
altered auditory feedback (AAF), and these sorts of things developing computational models.

333
00:33:46,880 --> 00:33:52,640
So those were the two places I could look for what might be going on computationally. And the odd

334
00:33:52,640 --> 00:33:58,560
thing that I discovered was that they seemed to be studying exactly the same thing even at the same

335
00:33:58,560 --> 00:34:03,760
level if you look closely. But they were not talking to each other assumed that they were studying

336
00:34:03,760 --> 00:34:11,440
different things and used completely different vocabularies. So, I spent a couple of years just going

337
00:34:11,440 --> 00:34:16,400
over all the motor control literature including visual motor stuff, going over psycholinguistic

338
00:34:16,400 --> 00:34:22,880
literature to see if I could somehow try to integrate these things. And I found that both have

339
00:34:22,880 --> 00:34:28,160
truths and that some form of integration was a worthwhile enterprise. And so that's how I

340
00:34:28,880 --> 00:34:35,520
started getting into motor control principles and integrating them with my ideas about how

341
00:34:35,520 --> 00:34:41,600
linguistic processes work. Yeah, so I mean I think it's a 2012 paper where you first

342
00:34:41,600 --> 00:34:48,160
lay out that model, right? And then it gets developed more in this book or maybe not in more detail

343
00:34:48,160 --> 00:34:54,000
but like maybe conceptually more. Yeah, it's kind of maybe organized a bit better. My first

344
00:34:54,000 --> 00:35:01,920
foray was actually with John Houde, who was a motor speech person at UCSF who was a fellow MIT

345
00:35:01,920 --> 00:35:07,520
student with me when I was there, David Poeppel and John Houde, we were there and John was doing

346
00:35:07,520 --> 00:35:12,480
this interesting motor speech stuff that I had no idea what it was. But we teamed up because I

347
00:35:12,480 --> 00:35:16,640
looked at John's motor speech models and tried to see if we could integrate them. So, we wrote a

348
00:35:16,640 --> 00:35:23,360
paper kind of arguing for some form of an integration. And then a 2012 paper extended it a bit and

349
00:35:23,360 --> 00:35:29,760
put it in more context. Yeah, but it is kind of laid out a bit more clearly I hope in the book.

350
00:35:29,760 --> 00:35:36,400
Yeah, I think it's more accessible. I mean, I remember reading the 2012 paper back in maybe 2012

351
00:35:36,400 --> 00:35:45,200
or so. And I liked it. But I feel like I understood it better now. Or although maybe that's just me

352
00:35:45,200 --> 00:35:53,200
growing up. But yeah, so it ends up being this kind of hybrid of a motor control approach and

353
00:35:53,200 --> 00:36:00,960
a psycholinguistic approach. Can you, do you think you can explain like how exactly you

354
00:36:00,960 --> 00:36:05,760
meld those things together? Like, yeah, those two different streams of research?

355
00:36:05,760 --> 00:36:12,000
Yeah, for sure. It's a fairly simple idea. So, motor control architectures basically have sensory

356
00:36:12,000 --> 00:36:19,200
targets. They have three parts, sensory targets, a motor execution or planning system, and then

357
00:36:19,200 --> 00:36:25,520
an in between translation system. So, you have these three components. And then you have computational

358
00:36:25,520 --> 00:36:31,120
operations like when you're motor planning, you can check to see whether the plans that you're

359
00:36:31,120 --> 00:36:35,760
developing will match the targets. And if they're not, you can do some error correction process. This

360
00:36:35,760 --> 00:36:44,400
is called feedback control. And there's internal and external forms of that as well. So basically the

361
00:36:44,400 --> 00:36:50,960
idea and in psycholinguistics, you have models that propose linguistic levels. Like if you look at

362
00:36:50,960 --> 00:36:56,560
Dell's model, you have a semantic layer of processing, you have a word level or a lemma level of

363
00:36:56,560 --> 00:37:02,400
processing. And then you have a phonological level of processing. So, three stages of processing.

364
00:37:02,400 --> 00:37:10,160
The idea that I had was that maybe each level of processing, like focusing on the phonological

365
00:37:10,160 --> 00:37:16,080
first, maybe the phonological level of processing in psycholinguistic models is actually

366
00:37:16,080 --> 00:37:24,800
composed of three parts. Like a motor control architecture, it has a phonological target system

367
00:37:24,800 --> 00:37:29,680
that's more related to sensory systems. It has a phonological motor system that's planning at

368
00:37:29,680 --> 00:37:35,440
the phonological level to hit those targets once executed. And then you have a translation system

369
00:37:35,440 --> 00:37:41,760
in between. And so that's basically the idea. That's what I proposed for how phonology looks.

370
00:37:41,760 --> 00:37:48,400
It's not just one box of units. It's three different boxes that are doing different things for

371
00:37:48,400 --> 00:37:57,280
phonological output planning. So, what else apart from phonology gets that sort of tri-partite

372
00:37:57,280 --> 00:38:02,720
division in your model? Yeah, so I started with phonology, but of course I was always thinking if

373
00:38:02,720 --> 00:38:09,760
you look back at the models that I drew, like in that 2012 paper, I have this three part division

374
00:38:09,760 --> 00:38:15,520
of labor in multiple, in a hierarchy. So, there's a low level of kind of phonetic control which

375
00:38:15,520 --> 00:38:21,280
involves low level motor cortex, low level somatic sensory cortex on the sensory side,

376
00:38:21,280 --> 00:38:26,880
and the cerebellum as the in between translation system. And then you go up a level and you're in

377
00:38:26,880 --> 00:38:32,080
the part of the, the dorsal stream, traditional dorsal stream as we proposed it. So

378
00:38:32,080 --> 00:38:40,080
posterior STG, inferior frontal regions, and then SPT in between. And then I drew a single box

379
00:38:40,080 --> 00:38:46,240
for the lemma level, the word level that fed into these systems. And as I was drawing this, I thought

380
00:38:46,240 --> 00:38:51,120
I wonder if this can be divided or should be divided into two as well, three parts as well.

381
00:38:51,120 --> 00:38:59,040
Knowing full well that people doing research on the word level have identified both anterior

382
00:38:59,040 --> 00:39:05,120
and posterior regions involved in word selection and planning, but there wasn't enough evidence

383
00:39:05,120 --> 00:39:12,480
for me to separate it out yet. So that eventually led to a collaboration with my former student,

384
00:39:12,480 --> 00:39:20,640
William Matchin, to develop this idea of a sensory motor-like architecture for morphosyntax 

385
00:39:20,640 --> 00:39:27,680
And that was our 2020 paper that proposed that. That didn't have a middle part. It only had the

386
00:39:27,680 --> 00:39:34,480
posterior side and anterior side. But just recently, and I think it just snuck into as a note in the

387
00:39:34,480 --> 00:39:41,840
book that we might have beyond to a in between area for morphosyntax. Where do you reckon that is?

388
00:39:41,840 --> 00:39:50,240
Inferior Parietal? Yes, of course. Saying that. So again, this is theorizing based on thinking about

389
00:39:50,240 --> 00:39:55,840
how the rest of the brain works, right? You have temporal lobe, coding, sensory stuff for vision,

390
00:39:55,840 --> 00:40:02,560
for audition. You have frontal lobe, coding, motor stuff. And the parietal lobe is the seat for

391
00:40:02,560 --> 00:40:07,600
visual motor transformation. It's where SPT lives for what we think is auditory motor translation.

392
00:40:07,600 --> 00:40:14,160
So where would the morphosyntax translation system be? Probably in the parietal lobe. And,

393
00:40:14,160 --> 00:40:20,240
you know, I went back and looked at some old ideas from David Gow, who was talking about things that

394
00:40:20,240 --> 00:40:25,520
made no sense to me at the time. But looking back, I was like, wait, I think he was on to something

395
00:40:25,520 --> 00:40:32,080
here with this. And then really great work by Kathy Price, subdividing the supramarginal gyrus (SMG),

396
00:40:32,080 --> 00:40:37,440
and her group into different components. One of them that seems to be very word like,

397
00:40:37,440 --> 00:40:45,280
and then turning up facts like people with conduction aphasia who have damage in that general zone,

398
00:40:45,280 --> 00:40:48,800
sometimes have pragmatic speech. They always have pragmatic speech.

399
00:40:48,800 --> 00:40:54,880
Not all of them. Some of them also have comprehension problems. And yeah, it gets a little

400
00:40:54,880 --> 00:41:01,200
complicated. But we shouldn’t start talking about how clean or messy aphasia is because that will get

401
00:41:01,200 --> 00:41:08,400
out of hand. Yeah. And when I say always, I mean in general. But yes. Yeah, no, that makes sense. So,

402
00:41:08,400 --> 00:41:15,680
so just this, the overarching scheme is that all of these different layers of the speech language

403
00:41:15,680 --> 00:41:23,200
system involve frontal, parietal and temporal components that are respectively motoric,

404
00:41:24,800 --> 00:41:31,680
translational sensory motor, I guess, because I said parietal in the middle, didn't I? And then sensory

405
00:41:31,680 --> 00:41:39,520
or connecting to the conceptual system, I guess. Yeah. So it's, I get in trouble for saying stuff

406
00:41:39,520 --> 00:41:46,560
like this because the linguists want to say, oh, we're just reducing language to the structure of

407
00:41:46,560 --> 00:41:53,360
language to sensory motor systems. But what I'm not saying, and I try to hammer this in the book,

408
00:41:53,680 --> 00:41:59,040
I'm not saying that this is just a sensory motor system. I'm saying that this is, has all the

409
00:41:59,040 --> 00:42:06,080
richness that, you know, whatever linguistics decides or determines is in the system. It's just

410
00:42:06,080 --> 00:42:12,800
distributed over this kind of abstract sensory motor like architecture. Yeah. Which is actually

411
00:42:12,800 --> 00:42:18,560
completely orthogonal to the linguistic divisions, right? And it's interesting, right? Because then

412
00:42:18,560 --> 00:42:23,280
you end up basically saying, well, yeah, every aspect of language is going to be frontal, temporal

413
00:42:23,280 --> 00:42:28,240
and parietal. And that's, there's lots of evidence for that, right? From both imaging and aphasia.

414
00:42:28,240 --> 00:42:34,880
And it starts, and you start to see how someone like our mutual friend, Ev, can, can come to the

415
00:42:34,880 --> 00:42:41,200
perspective that it's like that the language network is a rather unitary kind of structure.

416
00:42:41,200 --> 00:42:47,600
And I think that both you and I would tend to not agree with her on that final conclusion,

417
00:42:47,600 --> 00:42:53,120
but you can kind of see like what you can see the conditions for the evidence that leads her to

418
00:42:53,120 --> 00:42:59,200
think that way, right? Absolutely. Yeah. Because all of the facets do kind of end up like,

419
00:42:59,200 --> 00:43:05,120
you know, having your anatomical substrates throughout the network.

420
00:43:05,120 --> 00:43:10,800
That's exactly right. Yeah. It involves the whole that whole system, language involves that

421
00:43:10,800 --> 00:43:16,800
whole system at each level. So yeah, depending on how you, you know, what paradigms you use and

422
00:43:16,800 --> 00:43:22,400
whatever data you're collecting, you're going to see multiple systems involved. Exactly. Yeah.

423
00:43:22,400 --> 00:43:27,280
It's a very interesting and new idea. And I think it's like, yeah, made,

424
00:43:27,280 --> 00:43:34,400
like the argument, I think, is nice and clear in this book. And so I think everybody should read it

425
00:43:34,400 --> 00:43:39,280
just for that reason alone. You also, you have a chapter, and I don't want to get into this in too

426
00:43:39,280 --> 00:43:45,600
much detail, but I can't help being really interested in your chapter on the parallel hierarchy

427
00:43:45,600 --> 00:43:49,200
of speech production areas in the frontal, the dorsal, ventral division.

428
00:43:49,200 --> 00:43:56,160
And what you call the dorsal, pre-central speech area, which obviously, you know, that I'm interested in.

429
00:43:56,160 --> 00:44:02,080
So, can you tell us about, you know, what are your views on the dorsal, pre-central speech area?

430
00:44:02,080 --> 00:44:10,800
Where is it? What's it for? What's it doing? Yeah. So this, this is interesting. I got

431
00:44:10,800 --> 00:44:17,040
interested in this area because in functional imaging studies and all my SPT auditory motor circuit

432
00:44:17,040 --> 00:44:22,720
mapping studies, we would always see this dorsal pre-motor area that is right at the back of the

433
00:44:22,720 --> 00:44:28,400
of the middle frontal gyrus. If you had back towards the middle frontal gyros and hop over

434
00:44:28,400 --> 00:44:34,320
into the next, the pre-central gyrus, it's right there. It overlaps area 55B, which is in the

435
00:44:34,320 --> 00:44:41,280
Glasser Atlas based on human connectome project parcellation, interestingly, and it constantly shows up.

436
00:44:41,280 --> 00:44:46,800
It is one of the strongest language activations you see in your own study, Stephen. It was the one

437
00:44:46,800 --> 00:44:52,640
that showed up in your sensory and motor area in the frontal of the ventral, Broca, traditional

438
00:44:52,640 --> 00:44:58,880
Broca’s area didn't show up much. So, it showed up in Kathy Price's early data, her paper on hearing

439
00:44:58,880 --> 00:45:04,080
and saying, your work, my work, keep showing up. And I, we always wondered what it was, and I remember

440
00:45:04,080 --> 00:45:10,240
a conversation when you first joined my lab for a brief time, and we sat in my office and tried

441
00:45:10,240 --> 00:45:16,400
to figure out what this dorsal area was doing unsuccessfully at the time, but we were, you know,

442
00:45:16,400 --> 00:45:20,560
interested in what it was doing, and I didn't understand what it was doing until just a couple

443
00:45:20,560 --> 00:45:29,440
years ago. So basically, the insights came from, so overall, you have what appears to be two hierarchies

444
00:45:29,440 --> 00:45:35,200
of motor control or motor planning. One is their traditional ventral one, which involves broca’s

445
00:45:35,200 --> 00:45:43,040
as area, and a lower ventral motor cortex, which also activates in my auditory motor mapping studies.

446
00:45:43,040 --> 00:45:51,200
But then you have this more dorsal one. One of the breakthrough things came from Eddie Chang's work

447
00:45:51,200 --> 00:45:57,920
at UCSF showing that that region showed, in intracranial recordings, showed a correlation with

448
00:45:58,640 --> 00:46:03,360
pitch height during vocalization. So, it had something to do with pitch. And interestingly,

449
00:46:03,360 --> 00:46:10,000
there's been a lot of great work on mapping laryngeal motor cortex in humans. We have two of them.

450
00:46:10,000 --> 00:46:15,200
One of them is dorsal sitting right near this area that Eddie mapped, and it shows up in all our

451
00:46:15,200 --> 00:46:20,880
studies of sensory motor processes and speech. So that suggested an anatomical, functional anatomic

452
00:46:20,880 --> 00:46:26,160
connection between this dorsal speech area and dorsal laryngeal motor cortex.

453
00:46:26,160 --> 00:46:37,360
And then work by my grad student, Jon Venizia, had identified this area as being particularly

454
00:46:37,360 --> 00:46:42,960
auditory in its response properties. It showed spectral temporal receptive fields that coded pitch.

455
00:46:42,960 --> 00:46:51,360
And so this started me thinking about this area being important for pitch control via the larynx.

456
00:46:52,160 --> 00:46:57,680
And then thinking about what higher level functions might also be involved because that same area

457
00:46:57,680 --> 00:47:00,800
or just anterior to that also seems to be getting, 

458
00:47:00,800 --> 00:47:06,320
implicating syntax at some level or another. And so, the thought was, well, maybe it's

459
00:47:06,320 --> 00:47:12,320
prosody because we know that prosody has something to do with syntax. And this, you have this

460
00:47:12,320 --> 00:47:19,760
dorsal, middle-frontal gyrus, dorsal pre-motor speech area, dorsal laryngeal motor cortex hierarchy

461
00:47:19,760 --> 00:47:30,560
that's involved in coordinating respiration and prosody for pitch and prosody control via the

462
00:47:30,560 --> 00:47:35,360
dorsal laryngeal motor cortex. So that's the separation. Prosody kind of gets parceled out of the

463
00:47:35,360 --> 00:47:41,920
speech motor control system. And the rest of it, the phonetic control is the more ventral circuit.

464
00:47:42,560 --> 00:47:49,200
Okay, so yeah, to summarize that it's sort of, it's related to laryngeal motor, well, it's a

465
00:47:49,200 --> 00:47:54,640
adjacent to laryngeal motor cortex or one of the parts of that. It seems to be involved in pitch

466
00:47:54,640 --> 00:48:00,560
and prosody in contrast to more ventral frontal speech motor areas that are more about articulation.

467
00:48:00,560 --> 00:48:10,000
Yeah, this is really interesting. This is actually how I met Eddie, right? So, I met Eddie in 2007

468
00:48:10,000 --> 00:48:16,960
and he'd seen our fMRI paper on this area. And he was seeing it in his ECoG recordings and

469
00:48:16,960 --> 00:48:22,960
just being quite perplexed by it. Because another thing which I don't think you talk about in the

470
00:48:22,960 --> 00:48:28,960
book is that it has a very fast auditory response, right? So, this is a little spot in pre-motor cortex

471
00:48:28,960 --> 00:48:36,160
that responds to auditory stimuli within 100 milliseconds. So, you only see that with ECoG, right?

472
00:48:36,160 --> 00:48:42,240
We don't see it with our tools that we use. But Eddie was very struck by that and just basically

473
00:48:42,240 --> 00:48:45,600
trying to figure out what it was and that's how we became friends and started, you know, gradually

474
00:48:45,600 --> 00:48:53,200
developing some collaborations. And you mentioned that you guys found evidence that it

475
00:48:53,200 --> 00:49:01,440
has sort of auditory representations rather than motor, yeah? And I know that you've seen Eddie's

476
00:49:01,440 --> 00:49:09,520
stuff on that too, right? Where they showed in 2016 that it basically patent like an auditory area

477
00:49:09,520 --> 00:49:15,360
rather than a motor area. And so, you know, isn't it just, isn't it? Does it like, why is there a

478
00:49:15,360 --> 00:49:19,840
patch of auditory cortex sitting up in the pre-central gyrus? I mean, do you ever thought about it from

479
00:49:19,840 --> 00:49:25,040
that perspective? I mean, is this just like, does that need to be there because of the centrality of

480
00:49:25,040 --> 00:49:31,840
these coordinate transforms that your model is all about? Yeah, so that's we're thinking as an

481
00:49:31,840 --> 00:49:37,920
evolutionary biologist helps a bit. So, if you look in Macaques, their dorsal stream projects up to that

482
00:49:37,920 --> 00:49:45,120
general region. And you think about it and it projects up to near the frontal eye fields,

483
00:49:45,120 --> 00:49:49,600
which is kind of a misnomer. Frontal eye fields are not just about eye control. They're about

484
00:49:49,600 --> 00:49:58,160
orienting and some people think of it as a general orienting response. So in Macaque and in us,

485
00:49:58,160 --> 00:50:07,840
the function of that pre-linguistics, pre-language was apparently auditory orienting. So, hearing

486
00:50:07,840 --> 00:50:12,400
sounds in space and orienting towards them, controlling head movements, eye movements, attention

487
00:50:12,400 --> 00:50:18,880
towards those systems. So that was essentially the frontal control area that was using

488
00:50:18,880 --> 00:50:23,840
lower level auditory information. And of course, if you think about auditory localization,

489
00:50:23,840 --> 00:50:29,040
you immediately think about interaural level and time differences to orient in the horizontal plane.

490
00:50:29,040 --> 00:50:36,400
But the cues that are useful for orienting in the vertical plane are spectral. So much more rich,

491
00:50:36,400 --> 00:50:42,000
much more of the kinds of things that you'd find coded in primary auditory cortex. So all of that

492
00:50:42,000 --> 00:50:48,400
acoustic information would be useful in a pre-motor area that was important for orienting.

493
00:50:48,400 --> 00:50:55,840
And the idea that I have, that's hard to test, but the idea is that we evolved this dorsal

494
00:50:55,840 --> 00:51:03,920
area to control voice pitch where you also need pitch feedback. Because that's where the relevant

495
00:51:03,920 --> 00:51:08,240
sensory information was coming in. It was the part of the brain that was getting the relevant

496
00:51:08,240 --> 00:51:16,800
information. And so that's the logical place if you're evolving a system to evolve control

497
00:51:16,800 --> 00:51:22,160
of vocal pitch. And apparently that's what happened in birds as well, who have vocal learning

498
00:51:22,160 --> 00:51:27,840
for bird song. So yeah, I thought about it. It has to do with orienting, I believe, in attention.

499
00:51:27,840 --> 00:51:33,680
Yeah, I have always thought it was like attention related. I mean, I think that the first really

500
00:51:33,680 --> 00:51:39,520
clear evidence for this, the existence of this area was actually from MEG studies, like

501
00:51:39,520 --> 00:51:46,160
mismatch negativity where they had this 100 millisecond MMN response. And that was in the late 90s.

502
00:51:46,160 --> 00:51:50,800
And that was always kind of a scribe to an attentional thing, as it's coming in these like oddball

503
00:51:50,800 --> 00:51:55,760
paradigms. So yeah, this is really interesting. And this is a great example of like how you're,

504
00:51:55,760 --> 00:52:02,000
you know, you're, you're thinking as always very grounded in this sort of evolutionary perspective.

505
00:52:02,000 --> 00:52:07,360
So yeah. Yeah, yeah, I do try it helps. Again, it's just another source of constraint. It's really hard

506
00:52:07,360 --> 00:52:13,280
to do this work, right? In language neuroscience. And we need all the constraints from all the fields

507
00:52:13,280 --> 00:52:18,560
we can get. So, you know, I draw from motor control to cycle linguistics to general neuroscience,

508
00:52:18,560 --> 00:52:24,720
to evolutionary biology, anything I can get my hands on to help narrow down the search space,

509
00:52:24,720 --> 00:52:31,120
essentially. Yeah. Well, I really enjoyed the chapter on the, what do you call it? The

510
00:52:31,120 --> 00:52:37,680
Dorsal Pre-central Speech Area. Yeah, I'm broke. Yeah. Yeah. Okay. So last topic,

511
00:52:40,160 --> 00:52:46,480
very interesting, like, let's get back to that, that ventral stream. And I, as I emailed you

512
00:52:46,480 --> 00:52:51,280
originally, I wanted to talk about laterality. But before we kind of, you know, to set the,

513
00:52:51,280 --> 00:52:57,040
to set the stage, you're talking about laterality, can you talk about like the series of processing

514
00:52:57,040 --> 00:53:04,320
steps that happen in the ventral stream, kind of maybe which, Henschen outlined a hundred years ago,

515
00:53:04,320 --> 00:53:09,360
and they have further refined yourself and others have further refined here. So, what's that

516
00:53:09,360 --> 00:53:15,440
sort of processing hierarchy in your view? Yeah. So there's roughly speaking and it's obviously

517
00:53:15,440 --> 00:53:20,640
more complicated than this. But the, the rough sketch is that early on in auditory cortex, say in

518
00:53:20,640 --> 00:53:26,240
Heschl’s gyrus, there's coding for spectrotemporal information. That is information that

519
00:53:26,240 --> 00:53:35,040
varies spectrally and frequency over time. And that gets, that gets translated or from that you can

520
00:53:35,040 --> 00:53:41,280
derive phonological level representations. And I don't think that, that there are phoneme

521
00:53:41,280 --> 00:53:46,720
representations and auditory cortex. I think there are demy syllables or little chunks of syllables,

522
00:53:46,720 --> 00:53:51,840
but that's a separate issue. But something phonological is happening in the lateral superior temporal

523
00:53:51,840 --> 00:53:57,840
gyrus and dorsal bank of the superior temporal sulcus. From there, that information at the phonological

524
00:53:57,840 --> 00:54:06,400
level gets input and you can derive some information about word level. And by word, I mean abstract word

525
00:54:06,400 --> 00:54:14,240
and psycholinguistic terms. It is referring to lemmas and linguistic terms I'm talking about morphosyntax.

526
00:54:14,240 --> 00:54:20,800
And that's middle temporal gyrus, kind of regions that are, is coding word level stuff. And then

527
00:54:20,800 --> 00:54:26,880
from there, you're mapping it out into the conceptual semantic system. So, so acoustic features,

528
00:54:26,880 --> 00:54:32,640
the phonological to word level, the three main stages, I would say that we have a pretty decent handle on.

529
00:54:32,640 --> 00:54:39,440
Okay. And there's a lot of evidence for that, right? And like, you know, you point out that, you know,

530
00:54:39,440 --> 00:54:47,040
Jeff Binder’s 2000 fMRI paper, kind of like provides this really nice picture of what Heschl had

531
00:54:47,040 --> 00:54:54,880
come to 80 years before. So, I think we probably mostly, I mean, well, actually we don't all agree

532
00:54:54,880 --> 00:55:00,560
because there is dispute over whether it's anteriorly directed or posteriorly directed. But let's just say

533
00:55:00,560 --> 00:55:10,080
it's, we think it's mostly posteriorly directed in between us. And that's the basic layout of the

534
00:55:10,080 --> 00:55:16,080
the ventral stream. Now, then there's this other interesting aspect that you proposed in your papers

535
00:55:16,080 --> 00:55:24,080
in the first decade of the century, whereby the ventral stream is to some extent, which you can

536
00:55:24,080 --> 00:55:32,000
state for us, some extent, bilateral. So, what are your current views on the bilaterality of this

537
00:55:32,000 --> 00:55:38,080
series of steps? My current view is that at the level of phonological, up to the level of phonological

538
00:55:38,080 --> 00:55:45,440
processing, most of us are perfectly symmetric in terms of our lateralization, which is a controversial

539
00:55:45,440 --> 00:55:53,600
idea and something that I didn't even believe until recently. So, this site, so that just to state the

540
00:55:53,600 --> 00:55:58,560
facts, beyond that, once you get to the level of recognizing words and getting to higher levels,

541
00:55:58,560 --> 00:56:06,080
it becomes a little bit more left-dominant. But still, at the level of the ability to recognize words

542
00:56:06,080 --> 00:56:14,480
and understand their meaning, that is mostly bilateral in most people. So, in symmetric in many.

543
00:56:14,480 --> 00:56:22,640
So that's a summary of what I believe now. Okay. So, I don't agree with that. But I know that like

544
00:56:22,640 --> 00:56:30,000
you like a good argument. So, what's the evidence that leads you to come to that view? Like what are

545
00:56:30,000 --> 00:56:36,240
the main things that really struck you that led to that position? Yeah. Well, it's very hard to find

546
00:56:36,240 --> 00:56:43,440
people with unilateral damage to the superior temporal gyrus and has significant single-word

547
00:56:43,440 --> 00:56:50,160
comprehension problems. And by that, I mean not comprehension problems where there's lots of

548
00:56:50,160 --> 00:56:55,120
semantic foils or other sorts of things. This basic kind of can you tell the difference between

549
00:56:55,120 --> 00:57:01,920
bear and bear when you're pointing to pictures? We did a large-scale study on this and found that

550
00:57:01,920 --> 00:57:09,280
only about well less than 10% of people in chronic stroke have significant deficits, where

551
00:57:09,280 --> 00:57:17,600
significant is below like 85% something like that. So that's the fact. So, you know, most people,

552
00:57:17,600 --> 00:57:22,640
if you look at the distribution, most people are at ceiling on tasks like that, even with complete

553
00:57:22,640 --> 00:57:30,080
destruction of the superior temporal gyrus on the left. So that's one bit. If it was more

554
00:57:30,080 --> 00:57:36,480
lateralized, if it was significantly lateralized, we would expect to see more deficits in more people.

555
00:57:36,480 --> 00:57:40,480
So that's one argument. The other argument is if you damage the systems bilaterally,

556
00:57:40,480 --> 00:57:44,960
you end up with word deafness, that's where the severe deafness syndrome comes in.

557
00:57:46,080 --> 00:57:52,480
We've also done this in WADA and in acute stroke and we still can't identify severe single-word

558
00:57:52,480 --> 00:57:59,680
receptive deficits on these tasks. I'm just interrupting from the future here briefly because we've

559
00:57:59,680 --> 00:58:04,720
forgotten to define WADA, which makes this a bit hard to follow otherwise. So, WADA means the WADA

560
00:58:04,720 --> 00:58:11,600
test and that's named after a Japanese neurosurgeon named Juhn Atsushi Wada, who invented this procedure

561
00:58:11,600 --> 00:58:17,280
in the late 40s. This is a procedure where you use a barbiturate like sodium amobarbital to

562
00:58:17,280 --> 00:58:22,400
anesthetize one hemisphere of the brain at a time and this is done prior to surgery to determine

563
00:58:22,400 --> 00:58:29,120
lateralization of language or other functions. So, the idea is you transiently take one hemisphere

564
00:58:29,120 --> 00:58:34,080
out of the action and you can see what the other hemisphere can do. So, Greg and I are going to talk

565
00:58:34,080 --> 00:58:38,080
about some studies that have been done using this procedure to look at language lateralization.

566
00:58:38,080 --> 00:58:43,840
Okay, let's get back to it. And then I went the next step for this book to look back at the WADA

567
00:58:43,840 --> 00:58:50,160
data that I had collected and look to see because overall if you look at whether if people have left

568
00:58:50,160 --> 00:58:54,800
hemisphere anesthesia, you put the left hemisphere to sleep, you ask if they can comprehend words,

569
00:58:54,800 --> 00:58:59,360
then you do it with the right hemisphere. There are worse than comprehending words when the left

570
00:58:59,360 --> 00:59:04,560
hemisphere is asleep on average compared to the right. So, there's some asymmetry there. I never

571
00:59:04,560 --> 00:59:10,000
did it by that. But then if you look at the distribution of that asymmetry, it turns out that more than

572
00:59:10,000 --> 00:59:16,720
half of the people are perfectly symmetric and it's only a smaller fraction of people who are

573
00:59:16,720 --> 00:59:22,560
much more left dominant. So, my view is that the left dominance that we see in the population when

574
00:59:22,560 --> 00:59:29,120
we do functional imaging studies or group level stroke studies is being driven by very small biases

575
00:59:29,120 --> 00:59:36,240
in most people are pretty symmetric. A few people are slightly left dominant and then a small fraction

576
00:59:36,240 --> 00:59:41,120
of people are strongly left dominant. You average all those together and you get a kind of left

577
00:59:41,120 --> 00:59:46,320
dominance. Okay. That's what I think is going on at the population and individual level. Okay.

578
00:59:46,320 --> 00:59:52,880
Thanks. Yeah, that's a great summary. So, I want to kind of think about it from like, okay, things we

579
00:59:52,880 --> 01:00:01,360
agree on is that we sort of ask the question that like, I think both of us think about the question

580
01:00:01,360 --> 01:00:05,120
the same way. Like when we're asking like, what does it mean to be a bilateral? We're asking like,

581
01:00:05,120 --> 01:00:11,520
what can the right hemisphere do if put to the test, right? Is that the way you see it? Like,

582
01:00:11,520 --> 01:00:18,080
when you say that we're bilateral or that most people are bilateral, what you mean is that

583
01:00:18,080 --> 01:00:21,360
take the left hemisphere out of the picture, the right hemisphere can still comprehend words.

584
01:00:22,080 --> 01:00:28,080
Yep. Okay. I think another thing that we would agree on is that comprehension is much more

585
01:00:28,080 --> 01:00:33,440
bilateral than production, right? So, you've never made claims about the bilateral,

586
01:00:33,440 --> 01:00:37,520
bilaterality of speech production. You think that the production system is pretty lateralized?

587
01:00:37,520 --> 01:00:42,480
I think, well, that's traditionally my view, but now I'm questioning it because we're taking

588
01:00:42,480 --> 01:00:47,360
average it is and we're collecting by a sample. So, I think it is lateralized. I believe that

589
01:00:48,800 --> 01:00:53,680
and more so than comprehension, but I think we're overestimating the degree of lateralization.

590
01:00:53,680 --> 01:00:58,640
Okay. But we think that there's a difference. Yes.

591
01:00:58,640 --> 01:01:05,600
So yes, and I would agree with that too because like, I mean, what you said initially, the fact that

592
01:01:05,600 --> 01:01:12,080
single-word deficits, like word comprehension deficits at the single-word level are quite rare

593
01:01:12,080 --> 01:01:16,960
in aphasia. That's, as you said, that is just a fact that we have to grapple with.

594
01:01:18,240 --> 01:01:24,240
What it means is debatable, but that I would say comprehension is much more bilateral than production.

595
01:01:24,240 --> 01:01:30,560
I think another thing that we would agree on is that different aspects of comprehension

596
01:01:30,560 --> 01:01:34,880
would differ in the extent to which they're bilateral. So, for instance, you said, like,

597
01:01:34,880 --> 01:01:39,840
that spectro-temporal stage, we would both agree is fully bilateral.

598
01:01:39,840 --> 01:01:44,480
And then when you get to mapping and then the phonological word form stage,

599
01:01:45,440 --> 01:01:50,160
you think it starts to be maybe a little bit lateralized in some people, whereas I sort of think it's

600
01:01:50,160 --> 01:01:54,080
more than that. And then when you get to the stage of mapping and under meanings, you think,

601
01:01:54,080 --> 01:01:58,160
oh yeah, there's it's a little bit lateralized in some people, maybe more so in others.

602
01:01:58,160 --> 01:02:03,600
And I think, well, that's pretty strongly lateralized in my view. But we both agree that there's like,

603
01:02:03,600 --> 01:02:08,720
as you go through the hierarchy, it changes, right? It becomes more lateralized.

604
01:02:09,440 --> 01:02:16,160
Agreed. Yep. And another thing that I think is interesting, you didn't quite say this in the book,

605
01:02:16,160 --> 01:02:20,960
but I'm pretty sure you would think the same as me. When you get central, like, you know, as you're

606
01:02:20,960 --> 01:02:24,400
going through the hierarchy, like we just said, if you get more lateralized, but when you get central,

607
01:02:24,400 --> 01:02:30,720
you actually become bilateral again, right? Because meaning is very bilateral.

608
01:02:30,720 --> 01:02:35,760
Yeah, central meaning, like once you get up to concepts and yeah,

609
01:02:37,200 --> 01:02:42,080
yeah, I honestly never thought too much about it, but obviously, yeah, if you're going to get to

610
01:02:42,080 --> 01:02:48,880
the phonological form and then get to the word meaning, then you've got to have some bilateral

611
01:02:48,880 --> 01:02:52,720
capacity. Although, well, I'm not saying you have to have bilateral capacity. I'm just saying,

612
01:02:52,720 --> 01:02:59,520
if you can get through that bottleneck of sound meaning, you'd then have a semantic representation,

613
01:02:59,520 --> 01:03:04,560
which is very bilateral, as we see in stuff like Alex Huth's work, where everything looks

614
01:03:04,560 --> 01:03:11,040
really symmetrical. So those are, I think, of the things that we agree on. And I think what we

615
01:03:11,040 --> 01:03:21,760
disagree on is the extent of laterality of those mid-range, you know, that kind of that middle

616
01:03:21,760 --> 01:03:27,440
of the system, like where it's getting, you know, not the acoustic analysis and not the central

617
01:03:27,440 --> 01:03:35,360
representation of meaning, but just that bottleneck. And I guess, like, I guess I want to just ask your

618
01:03:35,360 --> 01:03:39,760
opinion about, like some of the, I think, things that are a little bit difficult for your viewpoint.

619
01:03:39,760 --> 01:03:46,800
So, and things are difficult for my viewpoint, right? So, I think, difficult for my viewpoint,

620
01:03:46,800 --> 01:03:51,840
for instance, is the fact that a few people with aphasia have profound word comprehension deficits.

621
01:03:51,840 --> 01:03:56,800
Difficult for your viewpoint is the fact that any people with aphasia have any word comprehension

622
01:03:56,800 --> 01:04:03,520
deficits, right? Like, so, you know, I'm sure you've seen our paper in brain from a couple of years

623
01:04:03,520 --> 01:04:10,400
ago. And, you know, like a lot of our people with large MCA lesions or large temporoparietal lesions,

624
01:04:10,400 --> 01:04:18,080
they do have pretty profound word comprehension deficits, certainly acutely, and in many cases

625
01:04:18,080 --> 01:04:23,840
into one and three months, resolving over time, which is super interesting, but like, what do you,

626
01:04:23,840 --> 01:04:31,520
what do you make of that initial single word comprehension deficits that we do see in most of our people?

627
01:04:31,520 --> 01:04:36,320
Well, not, I would never say most. I'd say more than half of our people with really substantial

628
01:04:36,320 --> 01:04:44,080
left temporoparietal damage. Yeah. Yeah. Yes, your paper is probably one of my favorites in terms

629
01:04:44,080 --> 01:04:53,360
of testing these ideas. So, which is why I want to get some of that data. I'm working on it.

630
01:04:53,360 --> 01:05:01,280
I'm working on it. Yeah. So, initially, so if you have giant frontal lesions, I, you know, I don't,

631
01:05:01,280 --> 01:05:08,480
I don't know exactly. I mean, diaschisis is an issue, right? Maybe the right hemisphere is suppressed.

632
01:05:08,480 --> 01:05:14,880
Maybe, maybe there's some frontal things that are, you know, people are just having trouble with

633
01:05:14,880 --> 01:05:19,680
selection processes, something that is not about the recognition of the word.

634
01:05:23,040 --> 01:05:29,600
Yeah. I'm not sure. In some ways, I think the giant lesions are a little bit hard to evaluate

635
01:05:29,600 --> 01:05:37,360
in the acute stage because I, you know, imagine that a lot of things are going on with these

636
01:05:37,360 --> 01:05:42,640
different lesions and tell things resolve a bit. But you would know that better than me. So,

637
01:05:42,640 --> 01:05:50,400
yeah. So, that's, I mean, the acute stuff is a concern, which is why we did this, the acute study,

638
01:05:50,400 --> 01:05:54,320
where we did it more focally. We had a bunch of people with acute stroke

639
01:05:54,320 --> 01:06:02,880
measured their blood flow and their lesion acutely and then looked to see whether people with

640
01:06:02,880 --> 01:06:10,000
substantial damage to these areas we think, like the STG, basically it was superior middle-front

641
01:06:10,000 --> 01:06:15,200
middle temporal gyri. If you had a lot of damage there, independent of how big the lesion was

642
01:06:15,200 --> 01:06:22,160
and where everything else was. Do you have a lot of people with significant deficits? And the answer,

643
01:06:22,160 --> 01:06:28,720
in that study was more than half of people had no problem whatsoever, despite damage to the ROI

644
01:06:28,720 --> 01:06:33,920
that's supposed to be lateralized. And then you start getting a trail off of ability in terms of

645
01:06:33,920 --> 01:06:42,000
the distribution where everyone is above threshold and only about, again, 20% are having some clinically

646
01:06:42,000 --> 01:06:47,680
significant deficits at single word. So, yes, it does happen acutely, but it's a smaller fraction

647
01:06:47,680 --> 01:06:53,440
and the distribution, I think, is what's interesting. So, the way I've started thinking about is rather

648
01:06:53,440 --> 01:06:59,600
than debating about whether that counts as lateralization. We just measure the distribution of these things

649
01:06:59,600 --> 01:07:05,280
and say, well, this is the degree of lateralization. The modal tendency is symmetric or whatever.

650
01:07:05,280 --> 01:07:10,720
And then you have 15% of people who are going to be below this level and that is just an empirical

651
01:07:10,720 --> 01:07:15,840
measure of lateralization. Yes. Okay. So, that's actually another thing that we agree on that I forgot

652
01:07:15,840 --> 01:07:21,680
to put in my list of things we agree on, is that there is very substantial individual variability

653
01:07:21,680 --> 01:07:27,680
in the degree of bilaterality. And I think we disagree on what the distribution is. I think we

654
01:07:27,680 --> 01:07:34,160
disagree quite significantly because I would never think that bilaterality would be the norm,

655
01:07:34,160 --> 01:07:40,480
but I understand that you do. So, I mean, like, from my point of view, when I think about those acute

656
01:07:40,480 --> 01:07:48,640
data and yes, it's definitely not inconsistent with what we see. I think that when you talk

657
01:07:48,640 --> 01:07:53,280
about an, like, you know, damage to a superior temporal ROI or even if it's in the middle temporal,

658
01:07:53,280 --> 01:07:59,120
I think in most people, there's going to be a lot of residuals left temporal function that's still

659
01:07:59,120 --> 01:08:05,760
going to be available to interpret the acoustic information from the right hemisphere. If indeed,

660
01:08:05,760 --> 01:08:11,200
like, left, if left, if left Heschl is really out of the picture, you need to do an awful lot of

661
01:08:11,200 --> 01:08:18,960
damage to take away, like, potential left hemisphere substrates for interpreting that spectrotemporal

662
01:08:18,960 --> 01:08:25,040
information that can come in through the right, which we all agree on. And in general, I'd say with

663
01:08:25,040 --> 01:08:31,680
functional imaging of people with aphasia, which is sort of basically my thing, there's a few

664
01:08:31,680 --> 01:08:39,200
things that strike me again and again. And one of them is that in almost everybody, you will be shocked

665
01:08:39,200 --> 01:08:44,240
by how much residual functional activity there is even in people that have huge lesions. So, we'll

666
01:08:44,240 --> 01:08:48,240
take somebody that has like a huge MCA lesion and you just think, "Well, that's just wiped out

667
01:08:48,240 --> 01:08:52,240
the language network." You look at it on a, you know, there are acute DWI and you're like, "That person's

668
01:08:52,240 --> 01:08:57,760
never going to talk again." Then we bring them in and scan them at like one month, three months,

669
01:08:57,760 --> 01:09:05,520
12 months is our goal, even at one month, or maybe at three. You'll just see activation right up to

670
01:09:05,520 --> 01:09:10,240
the edge of the lesion, you know, like, and then the lesion tries so hard to destroy the language

671
01:09:10,240 --> 01:09:15,920
network, but there's always bits of it left and it's very, it's extremely rare to get somebody that

672
01:09:15,920 --> 01:09:20,560
literally has no left hemisphere language network because these people, honestly, they don't survive

673
01:09:20,560 --> 01:09:26,400
because that would involve an entire hemisphere being destroyed. So, I'm very struck by,

674
01:09:27,440 --> 01:09:30,560
you know, I'm not saying that the right hemisphere isn't playing a role in comprehension or

675
01:09:30,560 --> 01:09:37,280
doesn't have any comprehension abilities, but I think that you might be overestimating the lack of

676
01:09:37,280 --> 01:09:42,160
left hemisphere substrates in some of these people, even if they've got damage to those.

677
01:09:42,160 --> 01:09:46,720
Yeah. Left hemisphere regions. What do you make of the WADA data then?

678
01:09:46,720 --> 01:09:52,960
Yes. Okay. So, the WADA data, well, I think that when in your WADA study,

679
01:09:55,040 --> 01:10:00,880
it was, you were testing them for quite a while because you were doing semantic and

680
01:10:00,880 --> 01:10:07,600
Phonological distractors, and I think that they were, I don't think that the hemispheres

681
01:10:07,600 --> 01:10:13,840
were fully out of the picture for a much of the period that you were testing them.

682
01:10:13,840 --> 01:10:22,160
And so, I think of what I've seen in our clinical WADA data that we've not published, by the way,

683
01:10:22,160 --> 01:10:28,640
so this is just anecdotes and should be not given full, you know, really should be back to

684
01:10:28,640 --> 01:10:35,120
publication, but like what we see is pretty complete single word comprehension loss in most people

685
01:10:35,120 --> 01:10:40,800
under anesthesia of the left hemisphere. And so, like I said, we haven't published that,

686
01:10:40,800 --> 01:10:45,440
but that's what we've seen in our data. But what who has published it is this

687
01:10:46,160 --> 01:10:53,680
Risse et al. from the late 90s, and they have, you know, when they, that's there, you know, they

688
01:10:53,680 --> 01:11:00,240
basically say that most people have zero single word ability when the left hemisphere is

689
01:11:00,240 --> 01:11:04,320
anesthetized. And I know that's quite different to what you found. But I think that there's a

690
01:11:04,320 --> 01:11:09,760
timing difference, and I know that you don't, and I know that you have reasons to not think that,

691
01:11:09,760 --> 01:11:15,680
because you found grip strength, like you didn't see a correlation between grip strength

692
01:11:15,680 --> 01:11:19,760
and comprehension. Yeah. Right. Yeah. Yeah. I mean,

693
01:11:19,760 --> 01:11:25,840
every method has its weaknesses. And I suppose there could be some, some, you know,

694
01:11:25,840 --> 01:11:32,640
groggy left hemisphere recovery that is underlying that ability. But I mean, I would expect more

695
01:11:32,640 --> 01:11:39,040
more deficits and more people. But yeah, be …. I think the task is important too. I,

696
01:11:39,040 --> 01:11:44,800
you know, like if you look at WAB comprehension, and my colleagues in South Carolina,

697
01:11:44,800 --> 01:11:48,960
when I tell them that speech perception is bilateral, they laugh at me because they see

698
01:11:48,960 --> 01:11:54,240
patients all the time who have dense single word comprehension deficits on the WAB.

699
01:11:54,240 --> 01:11:59,600
But the WAB is multiple pictures. There's semantic categories, and that's kind of bias

700
01:11:59,600 --> 01:12:04,640
into the left hemisphere more. So, you know, I'd have to look at these other cases in your new study

701
01:12:04,640 --> 01:12:09,040
that sounds really interesting what the task is. Oh, yeah, whether it'll ever be published, no,

702
01:12:09,040 --> 01:12:14,640
I mean, it's clinical data. So, I'm, it's just what we've seen, and when we looked into it.

703
01:12:14,640 --> 01:12:21,120
Okay. Yeah. I wish it was, I wish it was a new study. But yeah, no, the Risse one is the one that I,

704
01:12:21,120 --> 01:12:25,040
but like, yeah, like it's question, like they don't really, it's like an old clinical paper, right,

705
01:12:25,040 --> 01:12:29,520
from 30 years ago. They're not really like language neuroscience people, and they don't really

706
01:12:29,520 --> 01:12:35,600
describe their comprehension test in, in as much detail as we would like to, to have. So yeah,

707
01:12:35,600 --> 01:12:40,720
no, I think it's like, I agree with, I mean, I conceptually I agree with you. Like obviously,

708
01:12:40,720 --> 01:12:45,840
if you, if it was true that you could have anaesthetize a hemisphere and have single word comprehension

709
01:12:45,840 --> 01:12:51,680
proceed normally, that would be unequivocal evidence that the right hemisphere can do it.

710
01:12:51,680 --> 01:13:00,560
I'm just not sure that that's quite been empirically shown yet. Okay. Fair enough. And coincidentally,

711
01:13:00,560 --> 01:13:06,560
I'm also continuing to study this because yeah, we haven't convinced you and others yet. So, I'm

712
01:13:06,560 --> 01:13:12,640
continuing, but yeah, I mean, the good thing is it's an empirical question. And honestly, I don't care

713
01:13:12,640 --> 01:13:20,240
whether it's 80% lateralized or 20% or zero. I honestly don't care. Yeah. But the fact is that like

714
01:13:20,240 --> 01:13:28,080
you know, predicting behavior, aphasic deficits from lesion location is notoriously impossible. Like

715
01:13:28,080 --> 01:13:34,720
you almost never, I mean, exaggerating a bit, but it's very hard to do very, very,

716
01:13:34,720 --> 01:13:42,560
lesion. No, Greg, I don't agree. I don't agree. Well, I mean, but I'll tell you what, you know what

717
01:13:42,560 --> 01:13:49,680
the hardest deficit to predict is word comprehension. Comprehension, yeah. So word, no, word comprehension.

718
01:13:49,680 --> 01:13:54,080
Sentence comprehension is easy to predict. Word comprehension is very hard to predict. So I would,

719
01:13:54,080 --> 01:13:58,640
so I would definitely grant you that, but I do think that we can make, we can do better on a lot

720
01:13:58,640 --> 01:14:05,760
of other aspects of language. Yeah. So, what do you, what have you guys got going on to look

721
01:14:05,760 --> 01:14:12,160
at, look further at this question? We're trying to get funding. Who knows if we'll get it to recruit

722
01:14:12,160 --> 01:14:18,480
an unbiased sample of people with strokes left and right, give them a battery of tests and see

723
01:14:18,480 --> 01:14:24,880
where their lesions are. And then do the reverse. So, what we do in in lesion symptom mapping, of course,

724
01:14:24,880 --> 01:14:31,840
is identify a behavioral deficit and look for where in the brain we can predict that or correlate

725
01:14:31,840 --> 01:14:37,840
damage with that deficit. I want to do the reverse and I want to see given an area that we think

726
01:14:37,840 --> 01:14:44,240
is doing something for a function, look to see whether how many people would damage that area,

727
01:14:44,240 --> 01:14:48,960
like complete damage to that area actually have the deficit we'd expect. We tried that with a

728
01:14:48,960 --> 01:14:55,520
proxy of speech in our data set. First mapping the lesion location of apraxia of speech, which is like,

729
01:14:55,520 --> 01:15:01,360
you know, sensory motor cortex. And then looking to see how many people who have complete damage

730
01:15:01,360 --> 01:15:10,240
or a near complete damage to that area have it. And most do 70%, but 30% don't. And so my thinking is

731
01:15:10,240 --> 01:15:16,160
that maybe some of that variance is due to people who have more bilateral organization.

732
01:15:16,160 --> 01:15:21,440
Oh, I completely agree. And like we see basically the same thing with apraxia of speech. I mean,

733
01:15:21,440 --> 01:15:27,360
like it depends on what area you're thinking of like pre-central kind of issue. Yeah. So, I'd say in our

734
01:15:27,360 --> 01:15:32,720
data set of the people that have damaged there, probably about half or a bit more than half have 

735
01:15:32,720 --> 01:15:38,880
apraxia of speech. And then a very significant minority simply don't. And including some people with

736
01:15:38,880 --> 01:15:42,720
massive lesions where it's really not plausible that it could be like an adjacent region or that you

737
01:15:42,720 --> 01:15:48,880
didn't hit the central thing. And so, and so I think I've kind of coming to the view that like actually

738
01:15:48,880 --> 01:15:56,080
there are people in whom speech motor control can be subserved by the right hemisphere. And like you

739
01:15:56,080 --> 01:16:02,720
said, 30% yeah, it could be about that. And I like the idea that you said before like the answers

740
01:16:02,720 --> 01:16:08,000
to these questions are not going to be like, oh speech perception is bilateral or not. It's going

741
01:16:08,000 --> 01:16:14,560
to be what's the distribution of the capacity of the non-dominant hemisphere across the population.

742
01:16:14,560 --> 01:16:18,720
And it's not going to be binary and it's not going to be the same for any individual. And that's

743
01:16:18,720 --> 01:16:23,200
what our answers are going to look like. They're going to look like distributions in unbiased samples.

744
01:16:23,200 --> 01:16:29,120
So yeah, we're also start recruiting right hemisphere people now. So hopefully we'll be able to add

745
01:16:29,120 --> 01:16:37,520
that to our left. Hemi people and have some, some answers that satisfy us all. But yeah,

746
01:16:37,520 --> 01:16:40,320
I don't I don't think the right hemisphere people have too many deficits.

747
01:16:40,320 --> 01:16:46,480
No, no, I know. Yeah, yeah, I don't we found subtle ones in our sample.

748
01:16:46,480 --> 01:16:51,920
As did Kathy and her group. Yeah, but pretty subtle, right? So, I don't know that I don't think we're

749
01:16:51,920 --> 01:16:55,680
going to be shocked by the right hemisphere data, but we I agree with you. We need to do the due

750
01:16:55,680 --> 01:17:01,360
diligence and recruit those people as well. Yeah, yeah. And the people with no aphasia.

751
01:17:01,360 --> 01:17:04,800
Oh, absolutely. So yeah, that's how you know, that's how our lab works. It's like your, your

752
01:17:04,800 --> 01:17:10,080
inclusion is based on your legion, not on whether you have aphasia or not. You have a left hemisphere

753
01:17:10,080 --> 01:17:18,000
legion. We're coming to see you. And good. And, and yes, the Lily who is the lead SLP in my lab

754
01:17:18,000 --> 01:17:24,560
is incredibly good at getting people to agree to be in our study. Like she takes it as a she takes

755
01:17:24,560 --> 01:17:29,360
it as a personal affront if anybody ever declines consent. And like you'll literally like it'll

756
01:17:29,360 --> 01:17:33,440
happen like once per year and you'll like, you know, lose sleep over it. And like why didn't they

757
01:17:33,440 --> 01:17:43,280
consent? Yeah. So fascinating questions. And I mean, you know, I think you you put forward a really

758
01:17:43,280 --> 01:17:50,240
strong hypothesis about the bilaterality of speech perception and comprehension. And I think

759
01:17:50,240 --> 01:17:57,120
this is a lot of truth to it. And and I think that the facts about, you know, the overwhelmingly good

760
01:17:57,120 --> 01:18:00,800
word comprehension in most people with aphasia is not one that should be like ignored.

761
01:18:01,440 --> 01:18:07,680
Oh, um, but yeah, we have to figure out the exact distribution of these phenomena.

762
01:18:07,680 --> 01:18:12,720
Yeah, I mean, we need to ask the different question is language left dominant or not?

763
01:18:12,720 --> 01:18:17,280
Is the wrong question is like you were saying, it's what is the distribution? And that's kind of

764
01:18:17,280 --> 01:18:23,600
very by function, very a lot in individuals. And that's just an empirical question that we that

765
01:18:23,600 --> 01:18:29,200
we haven't the field hasn't yet started hasn't yet answered definitively. You started in your

766
01:18:29,200 --> 01:18:35,840
paper. I think that was great. And I've tried, but there's a gaps in all of this. Yeah, no, there's

767
01:18:35,840 --> 01:18:41,120
absolutely gaps. Yeah, I'm looking forward to the next 10 years. I think we will, I think we'll get to

768
01:18:41,120 --> 01:18:46,000
the bottom of this in 10 years. Yeah, we will, we will, we'll be in complete agreement about whatever the

769
01:18:46,000 --> 01:18:52,880
whatever the answer is. Well, that might results with you just so we can I can get the other

770
01:18:52,880 --> 01:19:02,240
perspective. Yeah, or Julius or Argye or you know, or just about anyone else. Yeah. Sometimes I'm like,

771
01:19:02,240 --> 01:19:07,680
oh, I'm maybe in like too like, you know, fiesty with Greg and I'm like, no, you're you're hearing

772
01:19:07,680 --> 01:19:12,800
the same thing from your from your co-authors. So I'm not like too ashamed of it. And like I and like

773
01:19:12,800 --> 01:19:16,560
you know, like I said, you like a good argument. And like, so I could share with our listeners that

774
01:19:16,560 --> 01:19:23,200
yeah, like I was postdoc in your lab for about a year. And the way that it happened, the way that you

775
01:19:23,200 --> 01:19:29,600
invited me to be a postdoc was that you reviewed one of my papers and largely disagreed with it.

776
01:19:29,600 --> 01:19:34,160
But we're like, and then you just emailed me after it's like, I reviewed your paper. I said, I said

777
01:19:34,160 --> 01:19:39,680
except don't really agree with you, but would you like to do a postdoc? So, I think it's like,

778
01:19:39,680 --> 01:19:46,880
that's just the right like that's the right attitude to like, you know, um, finding out the truth,

779
01:19:46,880 --> 01:19:51,760
right? Is like you find somebody you don't agree with and you uh work with them. Yeah, totally.

780
01:19:51,760 --> 01:19:56,960
You need you need someone with a different opinion. So, I mean, there's nothing I have no stake in

781
01:19:56,960 --> 01:20:01,920
the outcome. I just accept for I want to know how it works. What's you know, it's if we if the

782
01:20:01,920 --> 01:20:07,040
motor system's involved in speech perception, so be it. Oh yeah. And well, it should be said that on

783
01:20:07,040 --> 01:20:18,320
the question of that paper, um, you, you were right. Yeah. Okay. Well, um, I've taken a lot of your time.

784
01:20:18,320 --> 01:20:25,920
And I think we've gone through a lot of the big concepts of the book. Um, I hope that all the listeners

785
01:20:25,920 --> 01:20:32,240
will want to read it. Um, I really enjoyed it. And yeah, congratulations on on on writing this.

786
01:20:33,120 --> 01:20:38,480
Yeah, really it's got I think it's going to be real seminal piece of work. Oh, thank you. I really

787
01:20:38,480 --> 01:20:43,680
appreciate that and your opinion means much. And I always enjoy talking to you and debating you and

788
01:20:43,680 --> 01:20:49,840
yeah, it's, it's a good time. Thank you so much. All right. We'll take care. Great to talk and, um,

789
01:20:49,840 --> 01:20:57,360
look forward to seeing the book in print. Coming soon. All right. Bye. Take care. Bye. Okay. Well, that's it for

790
01:20:57,360 --> 01:21:02,400
episode 35. Thank you, Greg, for the very fun conversation. I hope everyone enjoyed it as much as I

791
01:21:02,400 --> 01:21:09,040
did. I've linked Greg's book on the podcast website at langneurosci.org/podcast and in the show notes

792
01:21:09,040 --> 01:21:14,320
as well as a few of the other key papers we discussed. Like I said, I really highly recommend this book.

793
01:21:14,320 --> 01:21:18,800
Anyone who listens to this podcast is going to enjoy the book. All right. Thanks also to Marcia

794
01:21:18,800 --> 01:21:31,760
Petyt for editing the transcript of this episode and thank you all for listening. Bye for now. See you next time.