1
00:00:00,000 --> 00:00:09,640
Welcome to episode 33 of the Language Neuroscience Podcast.
2
00:00:09,640 --> 00:00:14,160
I am Stephen Wilson, at the University of Queensland, in Brisbane, Australia.
3
00:00:14,160 --> 00:00:17,600
After the last couple of episodes about the business of science, I'm excited today to
4
00:00:17,600 --> 00:00:21,440
get back to doing an episode about a paper on the neuroscience of language.
5
00:00:21,440 --> 00:00:26,800
I'm joined by Steve Politzer-Ahles and Bernard Jap to talk about their excellent paper,
6
00:00:26,800 --> 00:00:31,280
‘Can the mismatch negativity really be elicited by abstract linguistic contrasts?’, which
7
00:00:31,280 --> 00:00:36,440
just came out in Neurobiology of Language, 2024, Volume 5, Issue 4.
8
00:00:36,440 --> 00:00:39,520
As I'm sure I've mentioned before, I'm on the editorial board of this journal, and
9
00:00:39,520 --> 00:00:42,920
I think that it's a great journal that is taking many of the right steps to improve academic
10
00:00:42,920 --> 00:00:43,920
publishing.
11
00:00:43,920 --> 00:00:48,560
So, I've decided to record some episodes on new papers in the journal that catch my eye.
12
00:00:48,560 --> 00:00:53,120
And Steve and Bernard's paper, from the most recent issue caught my eye for many reasons,
13
00:00:53,120 --> 00:00:55,600
as it will become clear in our conversation.
14
00:00:55,600 --> 00:00:59,760
One quick note, I am actively recruiting for several funded PhD positions in my lab in
15
00:00:59,760 --> 00:01:00,760
Brisbane.
16
00:01:00,760 --> 00:01:05,840
If you or someone you know might be interested, please check out langneurosci.org/join.
17
00:01:05,840 --> 00:01:07,520
Okay, let's get to it.
18
00:01:07,520 --> 00:01:10,240
Hi Steve, hi Bernard, how are you guys doing today?
19
00:01:10,240 --> 00:01:11,880
Great, how about you?
20
00:01:11,880 --> 00:01:12,880
Pretty good.
21
00:01:12,880 --> 00:01:16,840
It's 9am and sunny in Brisbane, and how about for you guys, where are you at?
22
00:01:16,840 --> 00:01:20,040
Yeah, I'm in Hong Kong, I'm doing pretty well, thanks.
23
00:01:20,040 --> 00:01:22,520
Just fighting a bit of a cold, maybe, we can hear this.
24
00:01:22,520 --> 00:01:25,800
It's actually pretty early here at 7am in Hong Kong.
25
00:01:25,800 --> 00:01:29,360
Uh-huh, well, good morning and thanks for waking up early for me.
26
00:01:29,360 --> 00:01:36,240
Yeah, I'm in Kansas, so it's 6pm, starting to thunderstorm, so hopefully my power doesn't
27
00:01:36,240 --> 00:01:37,760
go out in the middle of this.
28
00:01:37,760 --> 00:01:41,680
Okay, well that would be kind of atmospheric if we had like a power failure or something,
29
00:01:41,680 --> 00:01:44,800
but I guess it kind of doesn't really work out for zoom.
30
00:01:44,800 --> 00:01:46,480
Okay, great.
31
00:01:46,480 --> 00:01:50,840
So, can you guys, can we start just by like learning a little bit about you guys?
32
00:01:50,840 --> 00:01:52,920
Because I haven't met either of you before.
33
00:01:52,920 --> 00:01:54,360
I guess I'll start with you Steve.
34
00:01:54,360 --> 00:01:58,120
Like, can you just tell me a bit about yourself?
35
00:01:58,120 --> 00:02:02,800
Like how did you wind up at this point in your life researching the neural processing of
36
00:02:02,800 --> 00:02:04,600
linguistic contrast?
37
00:02:04,600 --> 00:02:06,880
Oh, that's a good question.
38
00:02:06,880 --> 00:02:12,960
I mean, I didn't expect to be doing neuro, so like it, I guess it's weird to say that
39
00:02:12,960 --> 00:02:17,280
like I figured out my interest when I was in grad school, because ideally that's something
40
00:02:17,280 --> 00:02:18,280
you'd know beforehand.
41
00:02:18,280 --> 00:02:22,000
But like, I went to grad school for linguistics.
42
00:02:22,000 --> 00:02:26,600
I think I actually, in my application, I said I was going to do theoretical syntax, so
43
00:02:26,600 --> 00:02:28,120
that's what I thought I was interested in.
44
00:02:28,120 --> 00:02:32,400
I mean, I am interested in it, but I ended up doing neuro.
45
00:02:32,400 --> 00:02:37,800
I think a lot of it came from like experiences I had taking,
46
00:02:37,800 --> 00:02:43,160
I took a neuro linguistics class with Rob Fierentino, and that's, I think, what got me
47
00:02:43,160 --> 00:02:48,240
interested in this, partly because I was just really excited
48
00:02:48,240 --> 00:02:54,400
in like how, the feeling of like that we were kind of finding linguistic phenomena
49
00:02:54,400 --> 00:02:58,320
that we could use to answer puzzles and questions.
50
00:02:58,320 --> 00:03:03,360
So, like throughout that class, we would always be like, oh, here's this thing that happens,
51
00:03:03,360 --> 00:03:07,680
but like we don't know if it, you know, do people ignore this noun because it's far away
52
00:03:07,680 --> 00:03:11,760
from the verb or because it's like not in a c-command relationship with the verb?
53
00:03:11,760 --> 00:03:16,400
And then be like, oh, we can come up with this fun sentence where like it's equidistant,
54
00:03:16,400 --> 00:03:18,440
but this one's in c-command and this one isn't.
55
00:03:18,440 --> 00:03:23,840
So, I just really loved that approach of like we find a really concrete question and then
56
00:03:23,840 --> 00:03:27,880
use our knowledge of language to figure out how to answer it.
57
00:03:27,880 --> 00:03:32,840
And I think I was also really excited in that class because especially back when I was starting
58
00:03:32,840 --> 00:03:38,160
to learn it, there was so much like low hanging fruit in neurolinguistics that people would
59
00:03:38,160 --> 00:03:43,280
always ask a question like, okay, so we learned about this thing about the N400, but what
60
00:03:43,280 --> 00:03:45,960
if we tried it with speakers of this language?
61
00:03:45,960 --> 00:03:49,960
Like, what if we tried it with a concurrent working memory task or something?
62
00:03:49,960 --> 00:03:54,240
And the answer was almost always like, I don't know if anyone's done that.
63
00:03:54,240 --> 00:03:56,000
Maybe we could do it.
64
00:03:56,000 --> 00:03:59,920
And so, it just, it made me really feel like this was a field that like I had a place in
65
00:03:59,920 --> 00:04:04,600
and I could contribute to and that just made me really excited to do neuro and I think
66
00:04:04,600 --> 00:04:06,640
I've loved doing it ever since.
67
00:04:06,640 --> 00:04:08,440
Yeah, that's great.
68
00:04:08,440 --> 00:04:12,240
Where was it that you went to grad school?
69
00:04:12,240 --> 00:04:14,720
Actually, it was Kansas, which is where I'm teaching now.
70
00:04:14,720 --> 00:04:15,720
Okay.
71
00:04:15,720 --> 00:04:21,520
And I was gone for like 10 years in the middle, but I had a great experience there.
72
00:04:21,520 --> 00:04:25,400
And when I found that there was a position open there, I was like, well, I know I love that
73
00:04:25,400 --> 00:04:26,400
department.
74
00:04:26,400 --> 00:04:28,640
So, I'm really happy that I got to come back.
75
00:04:28,640 --> 00:04:29,640
Are you from Kansas?
76
00:04:29,640 --> 00:04:33,560
So, did you just move there for the grad school experience?
77
00:04:33,560 --> 00:04:34,560
Yeah.
78
00:04:34,560 --> 00:04:36,880
No, I'm from Pennsylvania.
79
00:04:36,880 --> 00:04:41,440
So through like college and grad school, I kept gradually moving west and then I guess
80
00:04:41,440 --> 00:04:46,240
I'd gone all the way back around to the west and ended up back here.
81
00:04:46,240 --> 00:04:48,360
Okay, that's great.
82
00:04:48,360 --> 00:04:51,120
And how about you, Bernard?
83
00:04:51,120 --> 00:04:53,840
What was your story to winding up in this place?
84
00:04:53,840 --> 00:05:00,560
So, I started with my undergraduate degree where I ended up in the English Department of
85
00:05:00,560 --> 00:05:06,600
University of Indonesia and we had these different streams of cultural studies and literature
86
00:05:06,600 --> 00:05:08,080
and linguistics.
87
00:05:08,080 --> 00:05:15,320
I was instantly hooked to linguistics, shout out to the late Mr. Didin who was an amazing
88
00:05:15,320 --> 00:05:18,040
professor and something just sort of clicked.
89
00:05:18,040 --> 00:05:24,920
And at one point the following year, I was already an adjunct for morphology and syntax
90
00:05:24,920 --> 00:05:30,120
for my juniors, which looking back, I'm just sort of, it's just a little bit hilarious.
91
00:05:30,120 --> 00:05:34,560
I probably don't know what I was talking about back then.
92
00:05:34,560 --> 00:05:35,560
Yeah.
93
00:05:35,560 --> 00:05:40,520
Anyways, after graduation, I was fortunate enough to get into these European scholarship
94
00:05:40,520 --> 00:05:41,760
programs.
95
00:05:41,760 --> 00:05:44,760
This is called EMCL.
96
00:05:44,760 --> 00:05:48,520
European Masters for Clinical Linguistics and there I met an aphasiologist.
97
00:05:48,520 --> 00:05:51,560
So, you probably know her, Roelien Bastiaanse.
98
00:05:51,560 --> 00:05:52,560
Yeah, right.
99
00:05:52,560 --> 00:05:54,880
I don't know her personally but know her work very well.
100
00:05:54,880 --> 00:05:55,880
Yeah.
101
00:05:55,880 --> 00:05:58,920
And it was my Masters and PhD thesis.
102
00:05:58,920 --> 00:06:03,400
So, I did my Masters and PhD in the Netherlands, and I started with aphasiology.
103
00:06:03,400 --> 00:06:09,360
So, we were dissecting animal brains inside the class and then talking about cases of double
104
00:06:09,360 --> 00:06:14,880
dissociation and the famous sort of brain injury cases.
105
00:06:14,880 --> 00:06:20,400
And I received quite a bit of experimental training during these programs with EEG and MRI
106
00:06:20,400 --> 00:06:21,960
and whatnot.
107
00:06:21,960 --> 00:06:30,040
A lot of my beginning was actually just washing people's hair as an intern in an EEG lab.
108
00:06:30,040 --> 00:06:32,240
But yeah, I was sort of hooked from there.
109
00:06:32,240 --> 00:06:38,040
And there's something sort of remarkable, but actually seeing the brain's response to
110
00:06:38,040 --> 00:06:42,320
language, sort of unfolding in real time.
111
00:06:42,320 --> 00:06:51,000
I was also sort of passionate given the amount of impact that you can make because for Indonesian,
112
00:06:51,000 --> 00:06:55,480
for example, we're missing quite a few things.
113
00:06:55,480 --> 00:07:02,000
For example, in both aphasia, I was also doing a bit of research on dyslexia.
114
00:07:02,000 --> 00:07:10,320
I found out during my PhD that we didn't even have a test for assessing dyslexia.
115
00:07:10,320 --> 00:07:12,480
So that was something that we worked on.
116
00:07:12,480 --> 00:07:17,000
Anyways, that was a long and winded way to say.
117
00:07:17,000 --> 00:07:18,000
Yeah.
118
00:07:18,000 --> 00:07:19,000
Yeah.
119
00:07:19,000 --> 00:07:21,000
Well, that's, no, that's, it's so true.
120
00:07:21,000 --> 00:07:24,720
Like, it's really become apparent to me in the last few years since we put out the
121
00:07:24,720 --> 00:07:25,720
quicker aphasia battery.
122
00:07:25,720 --> 00:07:30,160
And like, people have written to me to ask to translate it into many different languages
123
00:07:30,160 --> 00:07:31,160
in the world.
124
00:07:31,160 --> 00:07:36,440
Or I've just heard this story again and again, like there is no aphasia assessment in my language,
125
00:07:36,440 --> 00:07:39,160
you know, and because it's creative, comments, people can just do what they want with it.
126
00:07:39,160 --> 00:07:44,640
So, I appreciate what you're saying about like, you know, a major language like Indonesian,
127
00:07:44,640 --> 00:07:46,680
probably 200 million speakers, right?
128
00:07:46,680 --> 00:07:48,800
Or something in that ballpark.
129
00:07:48,800 --> 00:07:49,800
Is that right?
130
00:07:49,800 --> 00:07:51,480
100 million maybe.
131
00:07:51,480 --> 00:07:53,800
What is it like?
132
00:07:53,800 --> 00:07:56,800
It depends how you define speakers like.
133
00:07:56,800 --> 00:07:57,800
Yeah.
134
00:07:57,800 --> 00:07:59,800
And it depends on how you define dialect continue.
135
00:07:59,800 --> 00:08:02,360
Okay, you never should ask a linguist a simple question.
136
00:08:02,360 --> 00:08:03,360
Okay.
137
00:08:03,360 --> 00:08:10,720
But, yeah, but you know, it's a very major language and yet basic assessments will be lacking.
138
00:08:10,720 --> 00:08:12,880
So much work to be done.
139
00:08:12,880 --> 00:08:13,880
Cool.
140
00:08:13,880 --> 00:08:14,880
Okay.
141
00:08:14,880 --> 00:08:15,880
Well, thank you.
142
00:08:15,880 --> 00:08:18,680
So, as you know, I wrote to you guys about your paper that just came out in the journal,
143
00:08:18,680 --> 00:08:21,440
Neurobiology of Language.
144
00:08:21,440 --> 00:08:26,400
And I thought it looked really interesting, and I wanted to kind of focus on, you know,
145
00:08:26,400 --> 00:08:29,160
new papers from this journal.
146
00:08:29,160 --> 00:08:33,960
And it is about the mismatched negativity or MMN.
147
00:08:33,960 --> 00:08:41,480
So, can you start by telling our listeners about the MMN, like, what is it?
148
00:08:41,480 --> 00:08:43,000
Why is it interesting?
149
00:08:43,000 --> 00:08:44,000
Yeah.
150
00:08:44,000 --> 00:08:52,320
So, the MMN, is this brain response that happens when your brain notices the difference
151
00:08:52,320 --> 00:09:01,440
between one stimulus or category of stimuli and suddenly gets a different stimulus or
152
00:09:01,440 --> 00:09:04,640
category of stimuli that doesn't fit.
153
00:09:04,640 --> 00:09:09,160
And I said, I was very intentional in saying when your brain notices it rather than when
154
00:09:09,160 --> 00:09:13,680
you notice it, because what's cool about the MMN is that this can happen even when you're
155
00:09:13,680 --> 00:09:15,680
not paying attention.
156
00:09:15,680 --> 00:09:16,680
Right.
157
00:09:16,680 --> 00:09:21,280
So, this is the most typical way of getting MMN, is people will hear like, bah, bah, bah,
158
00:09:21,280 --> 00:09:25,160
bah, bah, bah, bah, bah, bah, dah, bah, bah, bah, dah.
159
00:09:25,160 --> 00:09:30,120
and so, whenever you get this dah, which doesn't fit in with the bahs that you got used to,
160
00:09:30,120 --> 00:09:32,080
you got the MMN.
161
00:09:32,080 --> 00:09:35,480
But the cool thing is, you know, this will happen when people aren't paying attention to
162
00:09:35,480 --> 00:09:36,480
them.
163
00:09:36,480 --> 00:09:40,480
So, like, typically you let them watch in nature documentary and they're just having bahs
164
00:09:40,480 --> 00:09:46,040
and dahs playing into their ear and their brain will get excited every time there's a dah,
165
00:09:46,040 --> 00:09:49,480
even though you told them, don't worry about those sounds.
166
00:09:49,480 --> 00:09:53,920
It can even happen when people are asleep, although I haven't done that myself, but there are
167
00:09:53,920 --> 00:09:56,080
reports of that in the literature.
168
00:09:56,080 --> 00:10:04,080
So, it's really thought to be the brain's, like response to noticing violations of patterns
169
00:10:04,080 --> 00:10:10,080
sub-attentively or without your conscious attention.
170
00:10:10,080 --> 00:10:14,320
And the other thing that's cool about it that we talk about more in the paper is that it
171
00:10:14,320 --> 00:10:20,000
doesn't have to be something as simple as bah, bah, bah, dah, for your brain to notice it.
172
00:10:20,000 --> 00:10:26,640
The brain can pick up on pretty complex patterns and notice when things deviate from them.
173
00:10:26,640 --> 00:10:32,680
So, there are other reports in the literature of things like, you know, not just one tone
174
00:10:32,680 --> 00:10:36,920
that's different from other tones, but you can have tones that are following a certain
175
00:10:36,920 --> 00:10:42,040
kind of sequence and then you get another tone that's out of the sequence and you can get
176
00:10:42,040 --> 00:10:43,640
the MMN.
177
00:10:43,640 --> 00:10:51,040
So, it's very impressive that like the brain can track so many things without attention or
178
00:10:51,040 --> 00:10:53,200
conscious awareness of it.
179
00:10:53,200 --> 00:10:54,200
Uh-huh.
180
00:10:54,200 --> 00:10:55,200
Cool.
181
00:10:55,200 --> 00:10:59,560
And then just so that we make sure that our terminology is all on the table for our listeners,
182
00:10:59,560 --> 00:11:03,360
you have the, you know, the terms that get thrown around a lot in your paper would be standards
183
00:11:03,360 --> 00:11:04,360
and deviants.
184
00:11:04,360 --> 00:11:06,360
Can you tell us what standards and deviants are?
185
00:11:06,360 --> 00:11:07,360
It doesn't.
186
00:11:07,360 --> 00:11:08,360
It sounds a bit judgmental. (laughter)
187
00:11:08,360 --> 00:11:09,360
Yeah.
188
00:11:09,360 --> 00:11:15,440
Yeah. So, the standards are the, the, in our case, sounds, although it could be something else,
189
00:11:15,440 --> 00:11:20,360
the standards are the stimuli that you're getting a lot of and the deviants, right, is not
190
00:11:20,360 --> 00:11:27,240
necessarily deviant people that they're the, the stimuli that are rare in this, in this paradigm.
191
00:11:27,240 --> 00:11:30,560
So, if a person's hearing, bah, bah, bah, bah, bah, dah, the
192
00:11:30,560 --> 00:11:35,480
bahs are the standards and the dahs are the deviants, but of course the standards don't have
193
00:11:35,480 --> 00:11:41,800
to be just one sound. You can have, you know, the standards can be a whole bunch of different
194
00:11:41,800 --> 00:11:47,560
high tones of varying heights and the deviant can be a tone that's lower than all of them.
195
00:11:47,560 --> 00:11:48,560
Right.
196
00:11:48,560 --> 00:11:53,320
So, the standards is a category of things that you're hearing, you're hearing this category
197
00:11:53,320 --> 00:11:56,480
a lot or seeing this category a lot or whatever.
198
00:11:56,480 --> 00:12:00,360
And the deviance is the category that you're not getting as much of.
199
00:12:00,360 --> 00:12:01,360
Great.
200
00:12:01,360 --> 00:12:02,360
Thanks.
201
00:12:02,360 --> 00:12:03,360
Yep.
202
00:12:03,360 --> 00:12:10,560
Sorry, just to add, I think it's also important that the deviants are rare enough and the standards
203
00:12:10,560 --> 00:12:11,920
are frequent enough.
204
00:12:11,920 --> 00:12:18,280
The ratio that we have on the paper, I think is 1:7 deviants, the standard ratio
205
00:12:18,280 --> 00:12:20,240
or about 15% to 85%.
206
00:12:20,240 --> 00:12:22,240
To get a reliable.
207
00:12:22,240 --> 00:12:23,240
Okay.
208
00:12:23,240 --> 00:12:24,240
Yeah.
209
00:12:24,240 --> 00:12:27,440
So, yeah, for deviants to be deviant, then it needs to be something that they're in opposition
210
00:12:27,440 --> 00:12:28,440
to.
211
00:12:28,440 --> 00:12:34,640
It's like when Blink-182 became like a major record label, like, you know, like alternative
212
00:12:34,640 --> 00:12:35,880
rock was over, right?
213
00:12:35,880 --> 00:12:40,240
I mean, alternative rock could only be alternative rock when there was mainstream rock.
214
00:12:40,240 --> 00:12:43,280
And when the only thing that there was, was, alternative rock, there was no alternative rock
215
00:12:43,280 --> 00:12:44,280
anymore.
216
00:12:44,280 --> 00:12:51,000
I've got to steal that explanation next time I teach MMN. (Laughter)
217
00:12:51,000 --> 00:12:54,800
And then, you know, I think that thing, Bernard pointed out is really important because
218
00:12:54,800 --> 00:13:00,840
it, like, there is research finding you only get the MMN when the deviants are deviant enough,
219
00:13:00,840 --> 00:13:01,840
right?
220
00:13:01,840 --> 00:13:03,520
Like 15% or less.
221
00:13:03,520 --> 00:13:09,400
And that can be interesting to tell you that these things, that the standards are that
222
00:13:09,400 --> 00:13:12,240
your brain considers them to be one category.
223
00:13:12,240 --> 00:13:16,640
Because if you have, well, it's not only that the deviants have to be 50% but the standards
224
00:13:16,640 --> 00:13:18,440
have to be like 80%.
225
00:13:18,440 --> 00:13:24,760
So, if you have 40% of one thing, 40% of another thing and 15% deviants, you won't get
226
00:13:24,760 --> 00:13:29,680
mismatched negativity for that because you don't have one thing that you have 80% of.
227
00:13:29,680 --> 00:13:34,440
So mismatched negativity is like, it's not just a cool thing that happens, but it's also
228
00:13:34,440 --> 00:13:39,760
a useful test for seeing like what things does your brain lump together into one category?
229
00:13:39,760 --> 00:13:40,760
Right.
230
00:13:40,760 --> 00:13:41,760
Yeah.
231
00:13:41,760 --> 00:13:44,560
So, I mean, that's kind of getting to my next question, which is like what's it useful for,
232
00:13:44,560 --> 00:13:45,560
right?
233
00:13:45,560 --> 00:13:49,760
And I guess that you use it fundamentally to probe the brain's categorization abilities.
234
00:13:49,760 --> 00:13:52,240
Is that a fair summary?
235
00:13:52,240 --> 00:13:53,240
Yeah.
236
00:13:53,240 --> 00:13:59,040
And I think the goal of our paper was to see, like going back to I said earlier then and
237
00:13:59,040 --> 00:14:02,480
can pick up on pretty complex patterns and things.
238
00:14:02,480 --> 00:14:09,280
We wanted to just see how complex and how abstract your brain can get in terms of like how
239
00:14:09,280 --> 00:14:11,080
it's categorizing things.
240
00:14:11,080 --> 00:14:12,800
Great.
241
00:14:12,800 --> 00:14:15,800
And you know, it's called the mismatched negativity.
242
00:14:15,800 --> 00:14:18,320
So obviously you've defined what mismatch means.
243
00:14:18,320 --> 00:14:20,720
Can you tell us like, what does it mean to be negativity?
244
00:14:20,720 --> 00:14:25,120
Like how is this measured, and you know, what's the timing of it?
245
00:14:25,120 --> 00:14:31,600
Like what it just tell us a little bit about the sort of neuro inside of this component.
246
00:14:31,600 --> 00:14:37,720
So, usually you measure it by, you record the brain response you get with those deviants
247
00:14:37,720 --> 00:14:42,000
and you subtract out the brain response you would get from the same things when they're
248
00:14:42,000 --> 00:14:44,080
not deviants.
249
00:14:44,080 --> 00:14:50,000
The way that is often done is by you can play bah, bah, bah, dah to
250
00:14:50,000 --> 00:14:51,000
people.
251
00:14:51,000 --> 00:14:53,120
So bahs are standards and dahs are deviants.
252
00:14:53,120 --> 00:14:56,720
But then you can also play dah, dah, dah, dah, bah.
253
00:14:56,720 --> 00:15:01,520
So, the dahs are standards now and bahs are deviants and they're the exact same physical
254
00:15:01,520 --> 00:15:03,000
stimuli.
255
00:15:03,000 --> 00:15:07,560
So, you can record what's the brain response you get when dahs is a deviant minus what's the
256
00:15:07,560 --> 00:15:11,360
brain response you get when dah is a standard.
257
00:15:11,360 --> 00:15:16,200
And so presumably any brain response related to that physical stimulus is the same in both
258
00:15:16,200 --> 00:15:17,800
and gets subtracted out.
259
00:15:17,800 --> 00:15:22,280
So, what's left over is the mismatched negativity and it's called mismatched negativity because
260
00:15:22,280 --> 00:15:23,960
that tends to be negative.
261
00:15:23,960 --> 00:15:29,360
So, the brain response you get to the deviant tends to be more negative than the brain response
262
00:15:29,360 --> 00:15:32,560
you get to the standard unless you're a little kid.
263
00:15:32,560 --> 00:15:36,640
Then it's called the mismatched response and sometimes it's positive instead of negative.
264
00:15:36,640 --> 00:15:40,800
I'm not a child EEG person so I don't know all the ins and outs of that.
265
00:15:40,800 --> 00:15:43,800
But for adults it tends to be a negativity.
266
00:15:43,800 --> 00:15:51,240
And it tends to happen pretty early, often around like 170, 200 milliseconds after you hear
267
00:15:51,240 --> 00:15:56,000
that sound or see that thing that's deviant.
268
00:15:56,000 --> 00:15:58,720
Although it can happen a little bit later.
269
00:15:58,720 --> 00:16:03,600
That's one of the conundrums or conundra that comes up in our paper.
270
00:16:03,600 --> 00:16:04,600
Yes, definitely.
271
00:16:04,600 --> 00:16:06,080
We're going to talk about that timing issue.
272
00:16:06,080 --> 00:16:11,600
But yeah, so typical around 170 to 200 milliseconds and you mentioned you're measuring
273
00:16:11,600 --> 00:16:13,480
with EEG.
274
00:16:13,480 --> 00:16:19,120
I think people also measure it sometimes with magnetoencephalography and call it the
275
00:16:19,120 --> 00:16:20,600
MMMN or something.
276
00:16:20,600 --> 00:16:24,520
But yeah, you guys are doing it with EEG, yeah.
277
00:16:24,520 --> 00:16:30,920
Yeah, I think MMF I've seen like the mismatched field since there's not necessarily a positive
278
00:16:30,920 --> 00:16:31,920
or negative.
279
00:16:31,920 --> 00:16:32,920
Okay, cool.
280
00:16:32,920 --> 00:16:38,240
So yeah, like you said, your paper is fundamentally about sort of categorization and what can
281
00:16:38,240 --> 00:16:47,000
be a deviant that's not just dependent on a simple physical stimulus, difference in
282
00:16:47,000 --> 00:16:48,320
the stimuli.
283
00:16:48,320 --> 00:16:54,120
And you give these really nice examples in your intro, which I was hoping you could share
284
00:16:54,120 --> 00:17:01,200
with our listeners where people's language experience really determines what counts as a
285
00:17:01,200 --> 00:17:04,080
standard, what counts as a deviant.
286
00:17:04,080 --> 00:17:09,760
And maybe you could tell us about the paper by Philips or Colin Phillips from 2000 in
287
00:17:09,760 --> 00:17:10,760
that respect.
288
00:17:10,760 --> 00:17:11,760
Okay.
289
00:17:11,760 --> 00:17:14,160
Yeah, that's a great one.
290
00:17:14,160 --> 00:17:20,320
Yeah, so that paper, so what was done in that paper was instead of just having bah, bah,
291
00:17:20,320 --> 00:17:26,800
bah, bah, bah, bah, dah,
292
00:17:26,800 --> 00:17:28,440
people would hear like, bah, dah, gah, bah dah, gah, gah, pah
293
00:17:28,440 --> 00:17:30,720
So, they had, oh, hang on.
294
00:17:30,720 --> 00:17:32,720
I think you're, the part of you thinking that.
295
00:17:32,720 --> 00:17:35,480
Okay, so yeah, this is an awkward thing in your paper, right?
296
00:17:35,480 --> 00:17:40,960
So, we've got Phillips et al., 2000, and the other Phillips et al., 2000.
297
00:17:40,960 --> 00:17:44,200
So, I'm thinking of the one, because I'm just kind of, I'm trying to build it from simple
298
00:17:44,200 --> 00:17:45,200
to more complex.
299
00:17:45,200 --> 00:17:47,720
I'm thinking of the one where they manipulate voice onset time.
300
00:17:47,720 --> 00:17:48,720
The different VOT.
301
00:17:48,720 --> 00:17:49,720
Yes, okay.
302
00:17:49,720 --> 00:17:55,200
Yes, that's the other, the bigger, the more famous Phillips et al. paper, I think.
303
00:17:55,200 --> 00:18:00,400
Yeah, but yeah, it's, so, yeah, this paper is a really amazing demonstration of that,
304
00:18:00,400 --> 00:18:05,760
like, yeah, you're, like you said, Stephen, that language background influences how your brain
305
00:18:05,760 --> 00:18:07,160
puts things into categories.
306
00:18:07,160 --> 00:18:13,000
So, you have, they had, da, da, da, da, da, da, da, da, da, da, ta, but then
307
00:18:13,000 --> 00:18:16,560
these das all have different voice on set times, right?
308
00:18:16,560 --> 00:18:21,760
So, the voice on set time is, you know, how long is this little puff of air between the
309
00:18:21,760 --> 00:18:26,240
end of your d and the beginning of your ah.
310
00:18:26,240 --> 00:18:31,200
And when that gets long, that tends to sound like a T to English speakers, and when it's
311
00:18:31,200 --> 00:18:37,800
short, it tends to sound like a D, but the exact, you know, where that cut off is, is, different
312
00:18:37,800 --> 00:18:40,440
for different people and in different languages.
313
00:18:40,440 --> 00:18:45,800
So, what they did was they had stimuli with a whole bunch of different VOTs, a bunch of
314
00:18:45,800 --> 00:18:47,560
different voice onset times.
315
00:18:47,560 --> 00:18:50,160
So, there was no one stimulus that was the standard.
316
00:18:50,160 --> 00:18:54,000
There were like a bunch of stimuli with a 10 millisecond, voice onset time, a bunch
317
00:18:54,000 --> 00:18:59,560
with a 15 millisecond, voice onset time, a bunch with a 20 millisecond and whatever.
318
00:18:59,560 --> 00:19:04,760
But if it turns out that like as an English speaker, my cut off is 30 milliseconds, like
319
00:19:04,760 --> 00:19:08,800
anything longer than 30 milliseconds sounds like a T to me and anything shorter sounds like
320
00:19:08,800 --> 00:19:15,040
a D, they took this like, this distribution of sounds with different voice onset times
321
00:19:15,040 --> 00:19:21,560
and put it on the side of that cut off so that like 85% of the ones you heard were ones
322
00:19:21,560 --> 00:19:27,120
that would sound like a D and 15% were ones that would sound like a T. And indeed people
323
00:19:27,120 --> 00:19:31,400
get a mismatched negativity for that, which is telling you that like your brain took all
324
00:19:31,400 --> 00:19:34,320
those different sounds and decided these are all T's.
325
00:19:34,320 --> 00:19:39,560
So even though they're physically different sounds, or no, so I decided they're all D's,
326
00:19:39,560 --> 00:19:40,560
right?
327
00:19:40,560 --> 00:19:44,800
So even though they're physically different sound, your brain lumps them into the D category,
328
00:19:44,800 --> 00:19:48,040
and that can be a standard and that gives you a mismatched negativity.
329
00:19:48,040 --> 00:19:53,280
Whereas in the second experiment, they took that whole distribution and just shifted everything
330
00:19:53,280 --> 00:19:54,480
up.
331
00:19:54,480 --> 00:19:59,800
So instead of 10 milliseconds, now it's 30 milliseconds or whatever.
332
00:19:59,800 --> 00:20:04,240
And so that made it so the cut off between D and T was right in the middle of the distribution
333
00:20:04,240 --> 00:20:05,240
now.
334
00:20:05,240 --> 00:20:09,800
So, they had the exact same variation of voice on set times, but now half of them sound
335
00:20:09,800 --> 00:20:14,920
like a D and half of them sound like a T and you don't get mismatched negativity anymore,
336
00:20:14,920 --> 00:20:19,440
which tells us that it's really not the physical variation in the sounds that's giving you
337
00:20:19,440 --> 00:20:21,280
this mismatched negativity.
338
00:20:21,280 --> 00:20:26,160
It's the fact that your brain is taking 80% of these sounds and thinking these are all
339
00:20:26,160 --> 00:20:30,120
D's and the other 15% are T's.
340
00:20:30,120 --> 00:20:32,000
That's what gives the mismatched negativity.
341
00:20:32,000 --> 00:20:35,360
It's really like how your brain lumps these things into categories.
342
00:20:35,360 --> 00:20:36,360
Cool.
343
00:20:36,360 --> 00:20:39,960
And that's because that's where the typical English speaker puts that boundary between those
344
00:20:39,960 --> 00:20:42,240
two phonemes.
345
00:20:42,240 --> 00:20:48,640
And so, you see that it's linguistic knowledge that is shaping the nature of this categorization
346
00:20:48,640 --> 00:20:49,640
process.
347
00:20:49,640 --> 00:20:54,240
And then you talk a little bit in your paper about, I'm not going to go through all 12 of
348
00:20:54,240 --> 00:20:59,320
the papers that you analyze in your intro, but I want to talk about one more, which is
349
00:20:59,320 --> 00:21:08,200
Kazanina et al., in 2006, where you really get to see the language specificity of this phenomenon.
350
00:21:08,200 --> 00:21:12,680
So, can you talk about what's so neat about that paper?
351
00:21:12,680 --> 00:21:13,680
Yeah.
352
00:21:13,680 --> 00:21:18,760
This is one of my favorite papers ever because it took that sort of finding.
353
00:21:18,760 --> 00:21:22,440
The same paradigm as the Phillips et al. paper we just talked about and then looked
354
00:21:22,440 --> 00:21:26,520
at how that varies across people with different language backgrounds.
355
00:21:26,520 --> 00:21:30,960
And they did essentially that same experiment we just talked about, but they did it with
356
00:21:30,960 --> 00:21:34,840
Russian speakers, I think, and with Korean speakers.
357
00:21:34,840 --> 00:21:40,800
And the reason that's interesting is because in Russian, D&T are different sounds in the
358
00:21:40,800 --> 00:21:42,400
rule system of Russia, right?
359
00:21:42,400 --> 00:21:45,760
Like linguists, we call them, they're different phonemes.
360
00:21:45,760 --> 00:21:52,480
Whereas in Korean, D&T are not different phonemes, they're just different versions of the
361
00:21:52,480 --> 00:21:53,480
same sound.
362
00:21:53,480 --> 00:21:58,720
So, if you, I can never remember my Korean, but like if one of them is, if you have a certain
363
00:21:58,720 --> 00:22:02,640
sound at the beginning of a word, it gets pronounced this way, the same sound in the
364
00:22:02,640 --> 00:22:08,000
middle of a word gets pronounced this other way, but like a Korean speaker wouldn't, they
365
00:22:08,000 --> 00:22:10,600
don't give you different words.
366
00:22:10,600 --> 00:22:16,960
It's kind of like for English, the sound M, what we call the M, we pronounce it by touching
367
00:22:16,960 --> 00:22:20,640
our lips together if it's in a word like mom, but we pronounce it by touching our lips
368
00:22:20,640 --> 00:22:23,360
to our teeth if it's in a word like symphony, right?
369
00:22:23,360 --> 00:22:25,320
So, we don't think of it as a different sound.
370
00:22:25,320 --> 00:22:30,480
In the same way, like in Korean, these two sounds are not considered different sounds,
371
00:22:30,480 --> 00:22:33,440
they're just different versions of the same sound.
372
00:22:33,440 --> 00:22:39,520
And so, what they did in this experiment was they did that Phillips et al. experiment,
373
00:22:39,520 --> 00:22:42,240
once with Russian speakers and once with Korean speakers.
374
00:22:42,240 --> 00:22:47,760
And you found that with Russian speakers, there was a mismatch negativity because they
375
00:22:47,760 --> 00:22:50,120
sort these into two different categories.
376
00:22:50,120 --> 00:22:55,240
And in Korean speakers, there wasn't because they don't sort them into different categories
377
00:22:55,240 --> 00:22:56,240
along those lines.
378
00:22:56,240 --> 00:22:59,120
Yeah, super cool.
379
00:22:59,120 --> 00:23:03,440
So that really makes it clear that like linguist acknowledge plays a role in generating the
380
00:23:03,440 --> 00:23:08,360
MMN and shaping the categories that create the brain response.
381
00:23:08,360 --> 00:23:12,960
But your point in this paper is that even though they're clearly dependent on linguist
382
00:23:12,960 --> 00:23:17,880
knowledge, they're still ultimately generated by physical cues in the acoustic signal, right?
383
00:23:17,880 --> 00:23:22,880
So, in these cases, there's a change in voice onset time that either crosses a boundary
384
00:23:22,880 --> 00:23:28,640
or doesn't, but there's a physical change in the acoustic stimulus that drives that
385
00:23:28,640 --> 00:23:35,480
MMN and you kind of go through the whole literature and show that in all of the previous MMN studies,
386
00:23:35,480 --> 00:23:37,960
there's always a physical cue.
387
00:23:37,960 --> 00:23:44,480
And you are curious whether you can generate an MMN without any reliable physical cue.
388
00:23:44,480 --> 00:23:51,960
Can you tell us, tell me and our listeners, why was that an interesting question for you guys?
389
00:23:51,960 --> 00:23:58,200
Yeah, I think that it, I started wondering about that,
390
00:23:58,200 --> 00:24:03,200
especially when I was doing some MMN work with Kevin Schluter, who he and I were postdocs
391
00:24:03,200 --> 00:24:06,120
together at NYU Abu Dhabi at the time.
392
00:24:06,120 --> 00:24:11,960
And we were starting, like Kevin in particular has had a background in theoretical phenology.
393
00:24:11,960 --> 00:24:18,320
And so, we were interested in how the MMN is related to these abstract things that phonologists
394
00:24:18,320 --> 00:24:19,960
care about.
395
00:24:19,960 --> 00:24:24,120
And so, we were learning this literature and all these different things that are argued
396
00:24:24,120 --> 00:24:29,080
to show that like the MMN is really sensitive to abstract contrast.
397
00:24:29,080 --> 00:24:33,720
And we just started to wonder like how abstract can things really get because we were noticing
398
00:24:33,720 --> 00:24:36,800
this concern that you just mentioned that.
399
00:24:36,800 --> 00:24:41,960
So even in the Russian Korean study we just talked about, it's like a beautiful demonstration
400
00:24:41,960 --> 00:24:45,920
that the MMN cares about language knowledge.
401
00:24:45,920 --> 00:24:51,760
But then we thought what is really showing is the MMN or your language knowledge shapes
402
00:24:51,760 --> 00:24:55,400
which physical cues you pay attention to and how.
403
00:24:55,400 --> 00:25:01,480
But we really wanted to know like can you really get an MMN for like a phonological contrast
404
00:25:01,480 --> 00:25:06,120
that's like just for the abstract phonological contrast, not for the physical contrast that
405
00:25:06,120 --> 00:25:08,120
comes along with that.
406
00:25:08,120 --> 00:25:11,960
And as we kept looking into that, we kept finding like, oh, there's all of these things that
407
00:25:11,960 --> 00:25:16,800
look like they're very interesting abstract phonological contrast, but they're actually
408
00:25:16,800 --> 00:25:19,520
backed up by physical contrast.
409
00:25:19,520 --> 00:25:24,880
And there are also other interesting cases where you find that the MMN can get bigger or smaller
410
00:25:24,880 --> 00:25:28,720
as a result of interesting abstract linguistic things.
411
00:25:28,720 --> 00:25:34,840
So, like real words have bigger MMNs than non-words, phonologically underspecified sounds get
412
00:25:34,840 --> 00:25:35,840
smaller MMNs.
413
00:25:35,840 --> 00:25:37,880
So there are all these interesting things.
414
00:25:37,880 --> 00:25:42,320
But it's still like the physical contrast that's causing the MMN and then the interesting
415
00:25:42,320 --> 00:25:45,840
linguistic stuff that pushes it around to be bigger or smaller.
416
00:25:45,840 --> 00:25:51,400
So, we really just wanted to know like can you get an MMN for a just a purely linguistic
417
00:25:51,400 --> 00:25:52,400
contrast?
418
00:25:52,400 --> 00:25:54,480
All right.
419
00:25:54,480 --> 00:25:59,600
So, I'd like to ask you to describe your stimuli with your conditions.
420
00:25:59,600 --> 00:26:03,400
And I'm going to, I'd like to kind of do it in reverse order than how you do it in the
421
00:26:03,400 --> 00:26:05,960
paper because in the paper you kind of have your main condition that you're interested
422
00:26:05,960 --> 00:26:08,360
in and then you talk about your controls.
423
00:26:08,360 --> 00:26:12,280
But just because for in terms of explaining it to people that are like doing the dishes
424
00:26:12,280 --> 00:26:18,040
or going to work and I think it might be easier if we kind of build it up piece by piece from
425
00:26:18,040 --> 00:26:22,000
like the simplest condition to the most complex.
426
00:26:22,000 --> 00:26:30,200
So could you start by telling us, telling our listeners about your control condition that
427
00:26:30,200 --> 00:26:35,600
you call the aspiration contrast where it's the literally kind of the simplest?
428
00:26:35,600 --> 00:26:37,160
Yeah.
429
00:26:37,160 --> 00:26:40,920
Maybe I can take this one and Bernard can take over what we get to the more complicated
430
00:26:40,920 --> 00:26:45,000
ones that you'll probably remember the details of the stimuli better.
431
00:26:45,000 --> 00:26:47,360
But the control one was the simplest.
432
00:26:47,360 --> 00:26:54,440
This is just like the da da da da ta or ta ta ta da contrast where there really is just
433
00:26:54,440 --> 00:27:00,000
a physical difference between these like the ta has this extra puff of air or aspiration
434
00:27:00,000 --> 00:27:01,680
which da doesn't.
435
00:27:01,680 --> 00:27:05,640
And so, like you don't even need to know language to notice that these are different like
436
00:27:05,640 --> 00:27:09,000
chinchillas and things can get MMNs for these.
437
00:27:09,000 --> 00:27:14,320
So, this is we really had it there to make sure the experiment is working because like as
438
00:27:14,320 --> 00:27:18,000
we're going to see in a moment we had more complicated things where we weren't sure
439
00:27:18,000 --> 00:27:20,480
if we will get an MMN.
440
00:27:20,480 --> 00:27:25,760
And if you don't get an MMN in the complicated condition and you want to say that's interesting
441
00:27:25,760 --> 00:27:30,520
you need to be able to show that like we did the experiment right and we're able to get
442
00:27:30,520 --> 00:27:32,200
MMNs where you should.
443
00:27:32,200 --> 00:27:33,200
So, we have that.
444
00:27:33,200 --> 00:27:34,200
Okay.
445
00:27:34,200 --> 00:27:37,840
So, you got to be able to prove that you're capable of generating a standard MMN and so
446
00:27:37,840 --> 00:27:43,400
you do a very simple condition to establish that.
447
00:27:43,400 --> 00:27:49,640
And then you kind of have this middle of the road condition which is actually a replication
448
00:27:49,640 --> 00:27:51,840
of a study by Monahan et al., 2022,
449
00:27:51,840 --> 00:27:59,240
which is using a phonological category but it's still kind of driven by an acoustic
450
00:27:59,240 --> 00:28:00,240
cue.
451
00:28:00,240 --> 00:28:03,480
So, Bernard do you want to explain that one?
452
00:28:03,480 --> 00:28:04,480
Sure.
453
00:28:04,480 --> 00:28:09,520
So, in the sort of middle block, we have the voicing contrast.
454
00:28:09,520 --> 00:28:20,040
So, we have voiced phonemes like Bah, dah, Gah, vah, Zah and voiceless ones, Pah, Pah, Pah, Pah, Pah,
455
00:28:20,040 --> 00:28:25,880
and each of them act as deviants in one block and standards in another and then we just
456
00:28:25,880 --> 00:28:31,120
compare them as deviants and them as standards.
457
00:28:31,120 --> 00:28:40,120
So, for us to be able to elicit an MMN here it would mean the participants would have to
458
00:28:40,120 --> 00:28:41,120
be able to generalize.
459
00:28:41,120 --> 00:28:42,120
Yeah.
460
00:28:42,120 --> 00:28:46,960
And I'll add like the reason that this, we call this a little more abstract is because like
461
00:28:46,960 --> 00:28:51,760
the difference between voiced and voiceless is different among these.
462
00:28:51,760 --> 00:28:58,120
So, for the stops like Bah, dah, Gah, Pah Tah, Kah, the voiceless stop has aspiration,
463
00:28:58,120 --> 00:29:00,640
in it and the voiced stop doesn't.
464
00:29:00,640 --> 00:29:07,160
But for the fricatives like fah, sah, vah, zah, the difference between those is whether there's
465
00:29:07,160 --> 00:29:09,640
voicing during the fricative.
466
00:29:09,640 --> 00:29:16,440
So, the acoustic difference here is really different to the point that you don't have like 80%
467
00:29:16,440 --> 00:29:18,440
of your sounds all have a puff of air in them.
468
00:29:18,440 --> 00:29:20,280
That's not the case anymore.
469
00:29:20,280 --> 00:29:24,600
But phonologically these do make categories because like all the voiceless things do the
470
00:29:24,600 --> 00:29:30,360
same things like in English you have a different plural morpheme that goes after voiced versus
471
00:29:30,360 --> 00:29:31,360
voiceless.
472
00:29:31,360 --> 00:29:36,240
So, there's good reason to believe that these do make categories phonologically, but they don't
473
00:29:36,240 --> 00:29:40,920
have like 85%, 15% sort of category in acoustic space.
474
00:29:40,920 --> 00:29:41,920
Yeah.
475
00:29:41,920 --> 00:29:46,160
So, they make a really, so if you were to describe the rule and acoustic terms it would be very
476
00:29:46,160 --> 00:29:51,240
complicated because you'd have to have a different rule for stops and fricatives and that's
477
00:29:51,240 --> 00:29:56,480
what kind of makes it more abstract, although it's still ultimately a physical cue.
478
00:29:56,480 --> 00:30:02,040
You could describe a physical cue that distinguishes the standards and deviants.
479
00:30:02,040 --> 00:30:03,040
Yes.
480
00:30:03,040 --> 00:30:04,040
Okay.
481
00:30:04,040 --> 00:30:10,840
And then you kind of have your top of the pile most important critical condition where you
482
00:30:10,840 --> 00:30:17,480
do a contrast between verbs in the present tense and verbs in the past tense and you do
483
00:30:17,480 --> 00:30:23,720
this to try and create a situation where there are literally no systematic acoustic cues
484
00:30:23,720 --> 00:30:25,640
that could be generating the MMN.
485
00:30:25,640 --> 00:30:30,960
So, can you describe how this condition works and how did you guys come up with this one?
486
00:30:30,960 --> 00:30:31,960
Yeah.
487
00:30:31,960 --> 00:30:41,080
So we wanted to have a contrast with where the critical deviants’ standards have minimal
488
00:30:41,080 --> 00:30:42,880
physical contrast.
489
00:30:42,880 --> 00:30:46,400
It still does have a physical contrast because they're different words.
490
00:30:46,400 --> 00:30:53,200
But we have, for example, in the present tense we have ‘pave’, the verb ‘pave’, ‘get’ and
491
00:30:53,200 --> 00:31:01,600
‘Thank’ and the past sort of block would be ‘gave’, ‘met’ and ‘sank’.
492
00:31:01,600 --> 00:31:04,720
So, they only differ by their first sound.
493
00:31:04,720 --> 00:31:13,000
So, the way we sort of try to fool participants into thinking that this is just a series of words
494
00:31:13,000 --> 00:31:19,480
and they have to sort of abstract the actual tense feature is by also including a bunch of
495
00:31:19,480 --> 00:31:21,120
extra standards.
496
00:31:21,120 --> 00:31:32,280
So, among the block that features present tense, deviants we also featured past tense, deviants
497
00:31:32,280 --> 00:31:38,200
like toes, sang, blood, war and so on.
498
00:31:38,200 --> 00:31:45,280
So yeah, basically there's not really any systematic orthographic or chronological differences
499
00:31:45,280 --> 00:31:47,680
between the stimuli.
500
00:31:47,680 --> 00:31:52,640
I think we went through several iterations of this block because in English it's kind of
501
00:31:52,640 --> 00:32:00,160
difficult to find words that our only verbs, for example, or only nouns that have this sort
502
00:32:00,160 --> 00:32:01,800
of minimal contrast.
503
00:32:01,800 --> 00:32:09,320
And even then, I think in our stimuli we still saw that was like met, could mean different
504
00:32:09,320 --> 00:32:10,720
things to American speakers.
505
00:32:10,720 --> 00:32:14,880
For example, there isn't this, doesn't this mean the museum?
506
00:32:14,880 --> 00:32:16,880
Oh yeah.
507
00:32:16,880 --> 00:32:22,160
So that was also something that we were constantly thinking about.
508
00:32:22,160 --> 00:32:28,080
So, you had to find kind of culturally illiterate Hong Kong speakers of American English. (Laughter)
509
00:32:28,080 --> 00:32:32,280
Okay, so I'm just going to kind of restate this and just tell me if I get it right.
510
00:32:32,280 --> 00:32:38,760
So, like I think in the critical, in the critical direction you go with past tense standards
511
00:32:38,760 --> 00:32:41,400
and present tense deviants, is that correct?
512
00:32:41,400 --> 00:32:42,400
I think that's correct.
513
00:32:42,400 --> 00:32:46,600
I always have to remember which one is the, the under specified one.
514
00:32:46,600 --> 00:32:48,600
I think that's right.
515
00:32:48,600 --> 00:32:49,600
Yeah.
516
00:32:49,600 --> 00:32:53,120
Okay, so the standards would be, so I'm just reading off of your table.
517
00:32:53,120 --> 00:33:01,240
So, the standards would be something like ‘chose’, ‘sang’, ‘bled’, ‘swore’, ‘clung’, ‘plaid’, ‘grew’,
518
00:33:01,240 --> 00:33:10,120
‘drew’, ‘brought’, and then the deviants would be ‘pave’, ‘get’, ‘thank’.
519
00:33:10,120 --> 00:33:14,520
So, the deviants go into the present tense, and you match it perfectly phonologically by
520
00:33:14,520 --> 00:33:21,640
having the reversed version where the, where the deviants are ‘gave’, ‘met’ and ‘sank’, which have
521
00:33:21,640 --> 00:33:29,080
the identical coders to those critical stimuli but are past tense.
522
00:33:29,080 --> 00:33:33,080
And you need to of course use irregular English verbs, right?
523
00:33:33,080 --> 00:33:36,080
Can you explain why it had to be done that way?
524
00:33:36,080 --> 00:33:41,280
Yeah, because we wanted there to be no cue that tells you these are past, right?
525
00:33:41,280 --> 00:33:47,000
Because the goal was for people to hear like a long series of words where 85% of them are
526
00:33:47,000 --> 00:33:52,280
present tense and, and 15% are past tense and you have to figure out the tense to notice
527
00:33:52,280 --> 00:33:53,800
the difference.
528
00:33:53,800 --> 00:33:58,240
So, if we use regular verbs, then all of the past tense have, this ‘dah’ on the end and that would
529
00:33:58,240 --> 00:33:59,920
be a cue.
530
00:33:59,920 --> 00:34:05,840
So, using these irregulars, which like, you know, we're lucky that English is a weird language.
531
00:34:05,840 --> 00:34:12,920
So, we could have people hearing this like, Chose, sang, Bled, Swore, Get, Clung, Pled, Met,
532
00:34:12,920 --> 00:34:13,920
Thanked, right?
533
00:34:13,920 --> 00:34:18,800
So, it's, it's really hard to notice that they're all there like two past tense verbs sprinkled
534
00:34:18,800 --> 00:34:19,800
in there.
535
00:34:19,800 --> 00:34:24,880
So, it's a really different situation than like the, the control, VOT thing we started
536
00:34:24,880 --> 00:34:27,800
with, which is just like, Bah, Bah, Bah, Bah, Dah.
537
00:34:27,800 --> 00:34:32,960
It was really easy to, or I guess Bah, Bah, Bah, Bah, Bah, Pah, which was what it actually was.
538
00:34:32,960 --> 00:34:33,960
Yeah.
539
00:34:33,960 --> 00:34:42,760
We also wanted to keep it one syllable for each and some of the past tense English verbs
540
00:34:42,760 --> 00:34:43,760
would have two.
541
00:34:43,760 --> 00:34:44,760
Yeah.
542
00:34:44,760 --> 00:34:47,560
Sort of limited our option.
543
00:34:47,560 --> 00:34:51,440
So, yeah, you know, so you guys are really relying on this quirky property of English.
544
00:34:51,440 --> 00:34:54,560
So, I mean, do you think you could bond with Steven Pinker over this?
545
00:34:54,560 --> 00:34:58,640
So, over you interest in irregular verbs?
546
00:34:58,640 --> 00:35:02,240
Yeah, and then we'd have to , we'd have to compare regulars and irregulars.
547
00:35:02,240 --> 00:35:03,240
Oh yeah.
548
00:35:03,240 --> 00:35:06,640
If not comparing regular, irregular, he's not going to be interested.
549
00:35:06,640 --> 00:35:07,640
Yeah.
550
00:35:07,640 --> 00:35:13,080
So yeah, so you're really asking the brain to make a very abstract category here, right?
551
00:35:13,080 --> 00:35:16,120
Like you're asking it, and pre-attentively too, right?
552
00:35:16,120 --> 00:35:20,440
So, you're kind of even wondering whether this might happen if, when people are asleep even,
553
00:35:20,440 --> 00:35:27,640
like are people just pre-attentively extracting these syntactic features and to the point
554
00:35:27,640 --> 00:35:32,840
where the brain is capable of generating an MMN when the sequence is violated.
555
00:35:32,840 --> 00:35:35,640
And you're asking a lot of the brain.
556
00:35:35,640 --> 00:35:41,960
Yeah, and I mean, when we did it, we were like, it was like a long shot for this to even
557
00:35:41,960 --> 00:35:45,320
work because I mean, even if you're paying attention, right?
558
00:35:45,320 --> 00:35:53,560
If I say, "Choose, sing, bleed, swear, cling, met, plead, grow, woo, think, thank," right?
559
00:35:53,560 --> 00:35:57,120
It's really hard to notice that there's a past, present, contrast there.
560
00:35:57,120 --> 00:36:00,840
Like even for me having told you there was going to be there, like I have to read off
561
00:36:00,840 --> 00:36:03,560
the paper to do it. It's too hard for me to do it.
562
00:36:03,560 --> 00:36:09,040
So like to think that the brain could do that without paying attention, it's like very far
563
00:36:09,040 --> 00:36:10,040
fetched.
564
00:36:10,040 --> 00:36:15,080
So, we had to do a lot of methodological stuff to like, make sure we would be able to find
565
00:36:15,080 --> 00:36:20,840
an effect if it's there because we expected it to be really subtle if it exists at all.
566
00:36:20,840 --> 00:36:21,840
Right.
567
00:36:21,840 --> 00:36:26,960
Yeah, no, I mean, I'd be kind of surprised if it would come out.
568
00:36:26,960 --> 00:36:33,400
So, you pre-registered this experiment and I find that one of the really interesting aspects
569
00:36:33,400 --> 00:36:36,440
of this paper and I kind of wanted to talk to you guys about it.
570
00:36:36,440 --> 00:36:38,400
So, you didn't just pre-register it.
571
00:36:38,400 --> 00:36:42,480
You submitted it to neurobiology of language as a registered report, which means that you
572
00:36:42,480 --> 00:36:49,040
wrote the intro and methods and kind of like speculative results and what they would
573
00:36:49,040 --> 00:36:55,000
mean and submitted that for peer review before collecting the data.
574
00:36:55,000 --> 00:37:00,560
So, first, could you tell us like, why did you decide to take this tack with this paper?
575
00:37:00,560 --> 00:37:04,560
Yeah, so just a little bit of a context first.
576
00:37:04,560 --> 00:37:12,520
We submitted the registered report in 2021 and that was when I started as these postdoc
577
00:37:12,520 --> 00:37:18,240
and when I arrived in Hong Kong, he was still in Hong Kong, all the labs closed because
578
00:37:18,240 --> 00:37:19,880
of COVID.
579
00:37:19,880 --> 00:37:24,960
So we weren't, I arrived and then we weren't able to do any work at all.
580
00:37:24,960 --> 00:37:30,440
And yeah, so we discussed that this could be sort of a clever way to utilize our time to
581
00:37:30,440 --> 00:37:36,360
just write up maybe one or two registered reports and see how we can do.
582
00:37:36,360 --> 00:37:40,840
And of course, outside of the convenience factors, you know, to do something while the labs
583
00:37:40,840 --> 00:37:48,280
are all closed, it's sort of a way to commit to methodology before seeing any results.
584
00:37:48,280 --> 00:37:52,360
So, instead of the, as you said, instead of the traditional model where you run an experiment
585
00:37:52,360 --> 00:37:57,920
and let the results in submit, you sort of slip that process, you first develop your entire
586
00:37:57,920 --> 00:38:03,280
research plan, your methods, and so on, and sample size justification and submit that for
587
00:38:03,280 --> 00:38:04,600
peer review.
588
00:38:04,600 --> 00:38:10,200
And if you think that what you're proposing is solid and interesting, you might get an
589
00:38:10,200 --> 00:38:13,920
in principle acceptance before collecting any data.
590
00:38:13,920 --> 00:38:19,320
So, that is also an attractive feature of registered reports.
591
00:38:19,320 --> 00:38:25,600
So afterwards you get to run the study exactly as you plan and then you can submit sort
592
00:38:25,600 --> 00:38:30,920
of the second stage of the results for the final review.
593
00:38:30,920 --> 00:38:37,360
I also think that with studies that are as exploratory as this, we don't really know
594
00:38:37,360 --> 00:38:39,760
what to expect in terms of the findings.
595
00:38:39,760 --> 00:38:43,680
Actually, I'm not even sure when I was running the study that we would find the tense
596
00:38:43,680 --> 00:38:47,480
MMN, I think Steve also feels the same way.
597
00:38:47,480 --> 00:38:50,640
And if we didn't find it, would it be interesting enough, for example?
598
00:38:50,640 --> 00:38:52,720
You know, that's another question.
599
00:38:52,720 --> 00:38:54,720
Yeah.
600
00:38:54,720 --> 00:39:00,600
So, I do think this is sort of one of the biggest advantages.
601
00:39:00,600 --> 00:39:07,000
I think we've done two registered reports now and they've been quite interesting experiences.
602
00:39:07,000 --> 00:39:13,600
I also, if I can keep going here, I think we wanted to make sort of the best study
603
00:39:13,600 --> 00:39:21,080
ever by making sure our data is clean, and we have sort of strict inclusion criteria to the
604
00:39:21,080 --> 00:39:29,200
point where we were finding ways every day on the sort of on the road to submitting this
605
00:39:29,200 --> 00:39:33,880
registered report sort of how to make our lives more difficult by making it more and more
606
00:39:33,880 --> 00:39:34,880
big.
607
00:39:34,880 --> 00:39:43,480
I think at one point we had a few, sort of criteria to assess if we throw out a
608
00:39:43,480 --> 00:39:46,760
participant or not, they have too many bad channels.
609
00:39:46,760 --> 00:39:52,480
If we're excluding too much data during ICA and a few other factors.
610
00:39:52,480 --> 00:39:56,120
And then we were also trying to find out how do we do this objectively.
611
00:39:56,120 --> 00:40:04,000
So, we tried to do an analysis of the signal to noise ratio using a bootstrap method, which
612
00:40:04,000 --> 00:40:07,200
turned out to be a bit strict.
613
00:40:07,200 --> 00:40:08,200
Yeah.
614
00:40:08,200 --> 00:40:16,480
In any case, we also had a few cases, well, at least in this paper where we had to deviate
615
00:40:16,480 --> 00:40:18,520
from the original plan.
616
00:40:18,520 --> 00:40:19,520
Yeah.
617
00:40:19,520 --> 00:40:20,520
Okay.
618
00:40:20,520 --> 00:40:21,520
Yeah.
619
00:40:21,520 --> 00:40:22,520
Okay.
620
00:40:22,520 --> 00:40:23,520
Very, yeah.
621
00:40:23,520 --> 00:40:24,520
That's interesting.
622
00:40:24,520 --> 00:40:30,120
I was going to ask you about that because you, yes, things didn't go exactly as planned.
623
00:40:30,120 --> 00:40:35,000
But when you, so yeah, you've definitely, I've read your pre-registered version, and I see
624
00:40:35,000 --> 00:40:41,080
that it's identical to what you actually published, which is not always the case. (Laughter)
625
00:40:41,080 --> 00:40:45,720
But I'm curious like what the review process was like for the registered report.
626
00:40:45,720 --> 00:40:47,720
Like did you get a lot of feedback?
627
00:40:47,720 --> 00:40:51,240
Did you make any changes to your study plan based on the review process?
628
00:40:51,240 --> 00:40:52,240
Yeah.
629
00:40:52,240 --> 00:40:53,240
Absolutely.
630
00:40:53,240 --> 00:41:02,840
I think the first thing was the biggest sort of suggestion, one that we did not implement
631
00:41:02,840 --> 00:41:11,480
was the suggestion to split the experiment session into two because to have enough power,
632
00:41:11,480 --> 00:41:17,920
our study needed to be, just the experiment session needed to be around three hours.
633
00:41:17,920 --> 00:41:23,360
To around two hours, 45 minutes on average per person, participant, and then the editor,
634
00:41:23,360 --> 00:41:29,480
one of the editors said, this is too long, and you might want to split your session into
635
00:41:29,480 --> 00:41:31,720
two.
636
00:41:31,720 --> 00:41:38,040
We sort of pushed back on that because we know if we did that, there might be differences
637
00:41:38,040 --> 00:41:44,440
in small things like applying the cap and also the participants might not come back if
638
00:41:44,440 --> 00:41:46,200
we let them go.
639
00:41:46,200 --> 00:41:52,560
And I was kind of happy at the end to sort of keep it into just one session because this
640
00:41:52,560 --> 00:41:54,880
is quite long.
641
00:41:54,880 --> 00:41:55,880
And then we all had that.
642
00:41:55,880 --> 00:41:59,440
But there were changes that we did end up making, right?
643
00:41:59,440 --> 00:42:04,000
Some things got suggested, yeah, what were things about the words I remember?
644
00:42:04,000 --> 00:42:06,280
Stimuli.
645
00:42:06,280 --> 00:42:11,600
But yeah, I think, yeah, there were some words we had included that would have been problems.
646
00:42:11,600 --> 00:42:17,400
So, I think one of our present tense verbs was ‘run’ and a reviewer pointed out, oh, but
647
00:42:17,400 --> 00:42:19,920
‘run’, could also be of deverbal noun, right?
648
00:42:19,920 --> 00:42:25,560
I'm going to go for a run and that might break up your present tense verb category.
649
00:42:25,560 --> 00:42:30,800
So, we did a lot of like tweaks to the stimuli through that.
650
00:42:30,800 --> 00:42:37,200
And I think there are like some changes to the like interstimulus interval or things like
651
00:42:37,200 --> 00:42:38,200
that.
652
00:42:38,200 --> 00:42:46,240
So, a lot of sort of nitty-gritty changes, the experiment, which I was really like glad that
653
00:42:46,240 --> 00:42:50,840
we went through that process because it was so hard to find participants for this study.
654
00:42:50,840 --> 00:42:55,320
It would have been a real shame if we like spent a year getting participants and then submitting
655
00:42:55,320 --> 00:43:00,280
it and then submitted it and then found out like, oh, there was some fatal flaw, like a reviewer
656
00:43:00,280 --> 00:43:01,440
noticed.
657
00:43:01,440 --> 00:43:05,120
So, it was good to like get all those pointed out before we did the experiment.
658
00:43:05,120 --> 00:43:10,520
So, then we did the experiment in a way that we knew was like good and going to like not
659
00:43:10,520 --> 00:43:14,960
have problems that we hadn't thought of that other reviewers would have noticed.
660
00:43:14,960 --> 00:43:15,960
Right.
661
00:43:15,960 --> 00:43:22,960
I mean, yeah, ‘run’ is also like a point scored and cricket or you know, I think even the
662
00:43:22,960 --> 00:43:25,920
ladder in your stockings can be called a ‘run’.
663
00:43:25,920 --> 00:43:30,200
English is just so homophonous, you know, it's like, it's really difficult to design an
664
00:43:30,200 --> 00:43:32,200
unambiguous stimuli in English.
665
00:43:32,200 --> 00:43:33,200
Yeah.
666
00:43:33,200 --> 00:43:34,200
Okay.
667
00:43:34,200 --> 00:43:39,600
So, yeah, and you know, like I just think it's very admirable that you've kind of like cut
668
00:43:39,600 --> 00:43:43,200
yourselves off from the temptation to engage in p-hacking, right?
669
00:43:43,200 --> 00:43:47,840
Like you've released it and you're and this is a very detailed pre-registration I have to
670
00:43:47,840 --> 00:43:49,840
share with our listeners.
671
00:43:49,840 --> 00:43:53,720
Like, you know, it really gets it gets, all the methods are there.
672
00:43:53,720 --> 00:43:57,760
Like there's, there's not a whole lot of wiggle room in comparison to some others that I've
673
00:43:57,760 --> 00:44:00,400
seen where it's just like, we'll analyze the data with t-tests.
674
00:44:00,400 --> 00:44:02,920
It's like, what kind of t-tests?
675
00:44:02,920 --> 00:44:03,920
Yeah.
676
00:44:03,920 --> 00:44:09,480
And a lot of that is thanks to like, like Alec was the editor and the reviewers who anywhere
677
00:44:09,480 --> 00:44:13,080
where we weren't detailed, then they were like, you need to say exactly how you're going
678
00:44:13,080 --> 00:44:14,080
to do this.
679
00:44:14,080 --> 00:44:16,840
So, then we had to come back with a revision and say that.
680
00:44:16,840 --> 00:44:17,840
Yeah.
681
00:44:17,840 --> 00:44:22,920
In the context I assume you mean Alec Marantz, or co-author of some of these seminal
682
00:44:22,920 --> 00:44:23,920
MMN studies.
683
00:44:23,920 --> 00:44:27,680
Yeah, I mean, it's great to have a real expert overseeing that process.
684
00:44:27,680 --> 00:44:29,200
Okay, great.
685
00:44:29,200 --> 00:44:30,960
So, let's talk about what you guys found.
686
00:44:30,960 --> 00:44:35,000
And again, just kind of like from, let's go from the simplest condition and build it up
687
00:44:35,000 --> 00:44:36,000
from there.
688
00:44:36,000 --> 00:44:42,440
So, like, what did you see MMN wise for your aspiration contrast where it's just bah, bah, bah, bah, pah?
689
00:44:42,440 --> 00:44:45,560
Or maybe it was the other way around, I don't remember.
690
00:44:45,560 --> 00:44:52,520
Yeah, so and that one we got enormous MMN, thankfully, because if we didn't then we would
691
00:44:52,520 --> 00:44:55,640
have to think like we must have forgotten to plug in the cap or something.
692
00:44:55,640 --> 00:44:56,880
So yeah, 60 times. (Laughter)
693
00:44:56,880 --> 00:44:57,880
Yeah.
694
00:44:57,880 --> 00:45:02,080
It would happen, I guess.
695
00:45:02,080 --> 00:45:05,200
But yeah, so the like bah, bah, bah, bah, pah.
696
00:45:05,200 --> 00:45:09,280
No, I think it was the other way around, it was pah, pah, pah, pah, bah.
697
00:45:09,280 --> 00:45:20,440
So, the bah, gives you a way bigger, a way more negative VRP when it is a deviant than it is a standard and like in bah, bah, bah, pah.
698
00:45:20,440 --> 00:45:27,120
And that was around, you know, huge MMN and roughly around when and where it should be.
699
00:45:27,120 --> 00:45:33,080
It was so huge it was kind of all over the head, but it was also frontal which you often
700
00:45:33,080 --> 00:45:35,360
expect MMNs to be.
701
00:45:35,360 --> 00:45:38,640
And the timing was pretty much where you wanted it to be, right?
702
00:45:38,640 --> 00:45:46,200
I was about 300 milliseconds, maybe a little slower than some MMNs might be, but like definitely
703
00:45:46,200 --> 00:45:48,880
in the ballpark of expectations.
704
00:45:48,880 --> 00:45:49,880
Yeah, yeah.
705
00:45:49,880 --> 00:45:56,640
So, it's definitely a little bit slow, although it could also be like, you know, I'm trying
706
00:45:56,640 --> 00:45:59,400
to remember where we had actually time locked things too.
707
00:45:59,400 --> 00:46:03,120
So, if we had time locked things to like, we might have time locked to the beginning of the
708
00:46:03,120 --> 00:46:08,400
syllable and then, you know, in pah, like however many milliseconds later that you realize
709
00:46:08,400 --> 00:46:10,960
how long the aspiration is.
710
00:46:10,960 --> 00:46:13,160
But yeah, like, looking at the graph now,
711
00:46:13,160 --> 00:46:15,520
it's like a teeny bit before 300 milliseconds.
712
00:46:15,520 --> 00:46:20,680
So, it's later than the classic MMN, but it's still like within the ballpark of what I
713
00:46:20,680 --> 00:46:22,960
think you'd be willing to call an MMN.
714
00:46:22,960 --> 00:46:23,960
Yeah.
715
00:46:23,960 --> 00:46:24,960
And it's actually, I'm looking at it now.
716
00:46:24,960 --> 00:46:28,600
It's like definitely it's deviating clearly by 200.
717
00:46:28,600 --> 00:46:33,480
So, it's well on its way to being an MMN at that point.
718
00:46:33,480 --> 00:46:34,480
That's true. The huge effect is so
719
00:46:34,480 --> 00:46:35,480
Yeah.
720
00:46:35,480 --> 00:46:39,640
attention grabbing, that I look at like the peak latency, but I should actually be thinking
721
00:46:39,640 --> 00:46:41,240
of the onset latency.
722
00:46:41,240 --> 00:46:42,240
Right.
723
00:46:42,240 --> 00:46:43,240
Okay. Cool.
724
00:46:43,240 --> 00:46:44,240
725
00:46:44,240 --> 00:46:47,880
So, that one came out as expected, and that just kind of confirms that you're able to generate
726
00:46:47,880 --> 00:46:51,560
an MMN and everything's kind of in order.
727
00:46:51,560 --> 00:46:57,720
And then you next, let's look next at the middle contrast, which is the replication of Monahan
728
00:46:57,720 --> 00:47:05,040
et al., which is the sort of voicing violation where you've got standards that share a voicing
729
00:47:05,040 --> 00:47:10,280
category, and then the deviant is crosses that has the opposite value for that feature.
730
00:47:10,280 --> 00:47:12,880
So, what did you see for that one?
731
00:47:12,880 --> 00:47:13,880
Yeah.
732
00:47:13,880 --> 00:47:18,240
And there we, we still get an MMN, but it's now much weaker.
733
00:47:18,240 --> 00:47:20,040
Of course, I'm doing this.
734
00:47:20,040 --> 00:47:24,080
I always want a gesture with my hands, but I have to describe it verbally, right?
735
00:47:24,080 --> 00:47:30,280
But with the, with the, the, the, bah, bah, bah, pah, you have like an obvious huge spike
736
00:47:30,280 --> 00:47:31,280
here, right?
737
00:47:31,280 --> 00:47:36,480
One condition that has a giant peak sticking way up above the other ones.
738
00:47:36,480 --> 00:47:40,920
Here like both conditions are, you know, they're these lines and one is just a tiny bit higher
739
00:47:40,920 --> 00:47:42,720
than the other.
740
00:47:42,720 --> 00:47:47,880
So, we're getting an MMN, and it was statistically significant, according to all the tests that
741
00:47:47,880 --> 00:47:50,880
we pre-registered and said we were going to do.
742
00:47:50,880 --> 00:47:56,000
But it's much smaller like to the extent that if we haven't pre-registered it, like, you
743
00:47:56,000 --> 00:47:59,000
know, people might not trust is that really an effect there?
744
00:47:59,000 --> 00:48:00,000
Yeah.
745
00:48:00,000 --> 00:48:01,000
Okay.
746
00:48:01,000 --> 00:48:03,480
Let's see how the pre-registration is so key, right?
747
00:48:03,480 --> 00:48:07,200
Because there's like many different decision points there, like you could have had a different
748
00:48:07,200 --> 00:48:10,720
epoch, you could have had different plan stats.
749
00:48:10,720 --> 00:48:11,720
You're only looking at it.
750
00:48:11,720 --> 00:48:14,640
I mean, I don't want to get into this detail too much, but like you're only looking at one
751
00:48:14,640 --> 00:48:19,760
direction because of like the, the, the possibly that like markedness has a different effect
752
00:48:19,760 --> 00:48:20,760
than marked.
753
00:48:20,760 --> 00:48:25,520
And so like, you know, that, if that wasn't pre-specified as a reader, I'd be like, "Ah,
754
00:48:25,520 --> 00:48:26,520
that's interesting."
755
00:48:26,520 --> 00:48:27,720
That you decided to do that.
756
00:48:27,720 --> 00:48:30,280
But I know that it's there in advance.
757
00:48:30,280 --> 00:48:31,280
Okay.
758
00:48:31,280 --> 00:48:37,600
So, yeah, you've got this, you've got this effect that meets the criteria that you had set
759
00:48:37,600 --> 00:48:39,720
out in advance.
760
00:48:39,720 --> 00:48:45,840
And so, you've got the MMN and you're replicating that prior finding by Monahan and colleagues.
761
00:48:45,840 --> 00:48:46,840
Okay.
762
00:48:46,840 --> 00:48:53,040
And then finally, your critical manipulation of your novel tense contrast, what did you
763
00:48:53,040 --> 00:48:54,040
find for that one?
764
00:48:54,040 --> 00:48:57,840
Drum roll, like, we got MMN for that too.
765
00:48:57,840 --> 00:48:58,840
So, yeah.
766
00:48:58,840 --> 00:49:07,160
So, this is that like, give, pave, met, thank, sink, whatever, like somehow people, people's
767
00:49:07,160 --> 00:49:12,120
brains noticed when a past, when a present tense verb was thrown in among a bunch of past tense
768
00:49:12,120 --> 00:49:13,360
verbs.
769
00:49:13,360 --> 00:49:18,800
And there was a slight MMN there when they heard the present tense verb, which like again,
770
00:49:18,800 --> 00:49:21,560
it was a very weak MMN.
771
00:49:21,560 --> 00:49:27,200
Like, only, I can only live with myself calling it significant because we have that pre-registration
772
00:49:27,200 --> 00:49:30,040
otherwise, I don't know.
773
00:49:30,040 --> 00:49:32,680
But yeah, so we got this MMN.
774
00:49:32,680 --> 00:49:39,360
It was, I mean, actually, honestly, both this one and the phonological one were, they happened
775
00:49:39,360 --> 00:49:45,240
towards kind of the back of the head, which is not typically where you see MMNs come up.
776
00:49:45,240 --> 00:49:50,480
And they were also pretty late, like it was around 300-ish milliseconds that they were
777
00:49:50,480 --> 00:49:52,480
starting to happen.
778
00:49:52,480 --> 00:49:58,640
So, we got an effect, which was further back and later than MMNs tended to be, but it
779
00:49:58,640 --> 00:50:06,120
was elicited by the kind of contrast or by the kind of experiment design that is known
780
00:50:06,120 --> 00:50:07,920
to elicit MMNs.
781
00:50:07,920 --> 00:50:10,320
Yeah.
782
00:50:10,320 --> 00:50:11,320
Very cool.
783
00:50:11,320 --> 00:50:15,720
And I really think that the pre-registration was just so critical to enabling you to have
784
00:50:15,720 --> 00:50:19,000
that confidence.
785
00:50:19,000 --> 00:50:26,840
But as you said, the timing of the MMNs was a little unexpected, especially for the two
786
00:50:26,840 --> 00:50:29,600
more sort of speculative cases.
787
00:50:29,600 --> 00:50:34,040
And then you have this very nice plot in the paper that readers can listen, can look at
788
00:50:34,040 --> 00:50:39,920
if they want to, figure four, where you kind of show like which electrodes were significant
789
00:50:39,920 --> 00:50:42,600
when for the different conditions.
790
00:50:42,600 --> 00:50:47,120
And I think what's important is that the aspiration MMNs, so the basic, you know, the really
791
00:50:47,120 --> 00:50:53,520
kind of standard, easy one, kind of like everything's like significant about 200 milliseconds
792
00:50:53,520 --> 00:50:54,520
on, right?
793
00:50:54,520 --> 00:50:56,280
They start being significant.
794
00:50:56,280 --> 00:51:00,160
But for the other two, like for the voicing one, which is the second one we talked about,
795
00:51:00,160 --> 00:51:05,280
it really doesn't become significant until about 430 milliseconds.
796
00:51:05,280 --> 00:51:07,000
And then the tense one is actually a bit early than that.
797
00:51:07,000 --> 00:51:09,400
It's like about 330.
798
00:51:09,400 --> 00:51:12,960
So then later, and it's interesting because then you realize you go back and look at the
799
00:51:12,960 --> 00:51:17,040
plots and you realize, oh, yeah, that actual peak difference is not what's driving the
800
00:51:17,040 --> 00:51:21,320
significance for those more experimental conditions.
801
00:51:21,320 --> 00:51:27,240
It's really the later, like there's this sort of later longer lived, mushier negativity that
802
00:51:27,240 --> 00:51:29,320
actually, is driving the significance.
803
00:51:29,320 --> 00:51:31,800
So obviously you guys talk about all of them in the paper.
804
00:51:31,800 --> 00:51:37,360
I mean, how did that, how did that additional context like change how you felt about your
805
00:51:37,360 --> 00:51:39,360
results?
806
00:51:39,360 --> 00:51:40,360
Yeah.
807
00:51:40,360 --> 00:51:45,520
It was a little bit; it made us feel a little bit confused because once we were looking
808
00:51:45,520 --> 00:51:50,880
at them, our first reaction was like I just said, like so excited, yeah, we got an MMN.
809
00:51:50,880 --> 00:51:54,400
And then when we were looking at it more, we were like, wait, did we get an MMN? (Laughter)
810
00:51:54,400 --> 00:51:55,400
What is this?
811
00:51:55,400 --> 00:51:58,680
And then we thought, well, we had pre-registered this plan.
812
00:51:58,680 --> 00:52:01,760
If it comes out this way, we're going to call it an MMN.
813
00:52:01,760 --> 00:52:05,880
And then we thought in retrospect, maybe we should have said, like if all that happens
814
00:52:05,880 --> 00:52:10,840
and it's before 300, but we're like, well, that's not what we pre-registered and it wouldn't
815
00:52:10,840 --> 00:52:13,800
be fair to change our mind about things now.
816
00:52:13,800 --> 00:52:18,080
But we did have to have some thinking about is this really an MMN or is it something else
817
00:52:18,080 --> 00:52:23,880
happening later that we just happened to catch in our statistics?
818
00:52:23,880 --> 00:52:30,360
And we did end up, what we ended up arguing was that, well, we can nitpick about the timing
819
00:52:30,360 --> 00:52:37,840
and the place, but there's already existing research, including like, you know, Bornkessel-
820
00:52:37,840 --> 00:52:44,120
Schlesewsky’s model of like, predictive coding that argues that like, these same things might
821
00:52:44,120 --> 00:52:48,600
cause later or further back effects because it takes your brain longer to figure out the
822
00:52:48,600 --> 00:52:52,120
information that it needs to know this is a deviant.
823
00:52:52,120 --> 00:52:56,920
But we thought what's more important than the timing or the place is just the fact that
824
00:52:56,920 --> 00:53:01,880
this was elicited without people's attention because they, you know, we put them in a design
825
00:53:01,880 --> 00:53:05,560
where they're getting a bunch of past tense verbs and sometimes the present tense, we
826
00:53:05,560 --> 00:53:09,320
had them watch a nature documentary so they weren't paying attention to it.
827
00:53:09,320 --> 00:53:13,000
So, we figured like, that's really the hallmark of the MMN and more than anything else.
828
00:53:13,000 --> 00:53:16,880
It's like what, what kinds of things cause it to happen?
829
00:53:16,880 --> 00:53:20,480
And so here we felt that our is like, even if it's a little bit late and a little bit farther
830
00:53:20,480 --> 00:53:23,600
back, it's still caused by the things that are known to cause MMN.
831
00:53:23,600 --> 00:53:24,600
Right.
832
00:53:24,600 --> 00:53:25,600
Yeah.
833
00:53:25,600 --> 00:53:27,600
I mean, I, I, I see what you're saying.
834
00:53:27,600 --> 00:53:32,560
And I, I think that maybe the, thinking about the amount of processing that has to happen
835
00:53:32,560 --> 00:53:36,280
before you could legitimately get an MMN and it's, is maybe very key, right?
836
00:53:36,280 --> 00:53:41,080
Like, I mean, if you're just talking about like a really simple acoustic change, like, like
837
00:53:41,080 --> 00:53:46,760
hope, low and high tones, that information is going to make it to cortex in well less
838
00:53:46,760 --> 00:53:51,680
than a hundred milliseconds, at which point whatever changed detection mechanism can, can
839
00:53:51,680 --> 00:53:52,680
play out.
840
00:53:52,680 --> 00:54:00,440
But if you're talking like the syntactic one in particular, I mean, where's the, where's
841
00:54:00,440 --> 00:54:03,880
the disambiguation point of those words relative to other possible words?
842
00:54:03,880 --> 00:54:10,440
Like, at what point is that word able to be, you know, recognized as a present tense or
843
00:54:10,440 --> 00:54:12,200
past tense verb?
844
00:54:12,200 --> 00:54:14,560
It's certainly not at zero milliseconds.
845
00:54:14,560 --> 00:54:15,560
That's for sure.
846
00:54:15,560 --> 00:54:16,560
Right.
847
00:54:16,560 --> 00:54:17,840
It's, it's well into the syllable.
848
00:54:17,840 --> 00:54:22,800
And so is that inherently going to squish your MMNs later and, and probably spread them
849
00:54:22,800 --> 00:54:27,720
out a bit because it's probably not an identical disambiguation point across all your
850
00:54:27,720 --> 00:54:30,840
stimuli, even though its monosyllabic, right?
851
00:54:30,840 --> 00:54:31,840
Yeah.
852
00:54:31,840 --> 00:54:36,760
And so that's, I think, part of the reason why like our MMN for pah, pah, pah, bah, was really
853
00:54:36,760 --> 00:54:37,760
peaky.
854
00:54:37,760 --> 00:54:40,080
Like, there's a really sharp point where it happens.
855
00:54:40,080 --> 00:54:44,840
Whereas the ones for these other, they're kind of smeared out because, yeah, like, it's
856
00:54:44,840 --> 00:54:49,400
hard to know when exactly within that word, people are recognizing it.
857
00:54:49,400 --> 00:54:54,640
And so, in those situations, you get, you know, this is again, where I'm wanting to gesture,
858
00:54:54,640 --> 00:55:00,560
you get like a peak, a peak at 100 milliseconds and a peak at 120 and a peak at 130 and your
859
00:55:00,560 --> 00:55:06,320
average them together, they become a low smear across a long amount of time.
860
00:55:06,320 --> 00:55:11,400
But it's really hard to isolate when people are figuring that out.
861
00:55:11,400 --> 00:55:16,000
I had a, in a previous study, I tried like doing a, a gating task to figure out like, how
862
00:55:16,000 --> 00:55:20,400
many milliseconds into the word do people recognize which word it is?
863
00:55:20,400 --> 00:55:24,800
It's actually very, very hard and it varies across people.
864
00:55:24,800 --> 00:55:28,760
And even with this one, like, what we're really interested in is not when do they realize
865
00:55:28,760 --> 00:55:33,080
this is the word paved, but when do they realize this is a past tense verb?
866
00:55:33,080 --> 00:55:35,080
And that might be even different.
867
00:55:35,080 --> 00:55:39,800
So, we, that's another reason I think we really had to pre-register was because we were like,
868
00:55:39,800 --> 00:55:41,920
we don't have an easy solution to this problem.
869
00:55:41,920 --> 00:55:44,400
Let's just roll the dice and hope it works out.
870
00:55:44,400 --> 00:55:45,400
Yeah. (Laughter)
871
00:55:45,400 --> 00:55:47,400
So, we better pre-register it before we try that.
872
00:55:47,400 --> 00:55:51,840
Yeah. And you pre-registered, you basically said that you're going to look at the whole epoch
873
00:55:51,840 --> 00:55:58,600
and you just, and you described, I think, I think it was 700 milliseconds that you're going
874
00:55:58,600 --> 00:55:59,600
to look at.
875
00:55:59,600 --> 00:56:02,960
So yeah, you've kind of like committed yourself to that.
876
00:56:02,960 --> 00:56:07,720
And it's kind of fascinating, isn't it, that like even though you had a clear, even though
877
00:56:07,720 --> 00:56:12,080
you pre-registered and had a clear plan and you stuck to it, it still didn't really end
878
00:56:12,080 --> 00:56:14,520
up resolving the ambiguity of the data, right?
879
00:56:14,520 --> 00:56:17,600
The data, the data just never play nice, do they?
880
00:56:17,600 --> 00:56:22,760
They just never, they never really do what you wanted, even when you did everything right.
881
00:56:22,760 --> 00:56:23,760
No.
882
00:56:23,760 --> 00:56:27,240
That is so true.
883
00:56:27,240 --> 00:56:33,440
What did you think about this Bernard when the timing, when the timing was raised all these
884
00:56:33,440 --> 00:56:34,440
extra issues?
885
00:56:34,440 --> 00:56:41,240
Well, I'm just thinking about how to interpret them and sort of justify our interpretation.
886
00:56:41,240 --> 00:56:50,360
I think the negativity, especially in around the 400 millisecond mark is in the sort of
887
00:56:50,360 --> 00:56:52,600
central posterior area.
888
00:56:52,600 --> 00:56:58,880
If people don't know the context for this study, they might see this as an N400, but we know
889
00:56:58,880 --> 00:57:03,380
from several studies that N400 are modulated by attention.
890
00:57:03,380 --> 00:57:07,800
So, it doesn't really appear when people aren't paying attention.
891
00:57:07,800 --> 00:57:12,600
This is probably one of our sort of defense against it.
892
00:57:12,600 --> 00:57:21,640
We actually have another paper on this abstract mismatch negativity, and the negativity is also
893
00:57:21,640 --> 00:57:24,040
a bit later and coincides with the N400
894
00:57:24,040 --> 00:57:30,080
time window, and this is also again the way we interpret the component.
895
00:57:30,080 --> 00:57:34,760
Yeah, but you haven't manipulated attention, but I guess your people can't possibly be paying
896
00:57:34,760 --> 00:57:36,440
attention to this for three hours.
897
00:57:36,440 --> 00:57:43,720
I think you can just rely on human nature to guarantee that they were paying attention.
898
00:57:43,720 --> 00:57:50,040
That is one of my things I would like to do in the future is this and several other things
899
00:57:50,040 --> 00:57:54,520
that I've done with MMN where these questions have come up where I'm like, I'm not sure if
900
00:57:54,520 --> 00:57:59,320
this is actually MMN or if it's some other overlapping component that got smaller and
901
00:57:59,320 --> 00:58:01,680
made the MMM look bigger.
902
00:58:01,680 --> 00:58:05,640
That's something I've wanted to do is do some experiments where you do manipulate attention
903
00:58:05,640 --> 00:58:11,200
and have a block where they're told press the button every time you hear a past tense verb
904
00:58:11,200 --> 00:58:14,120
and another one where they're just watching a movie.
905
00:58:14,120 --> 00:58:18,040
Of course, the challenge is it took three and a half hours to do it this way.
906
00:58:18,040 --> 00:58:21,600
Am I going to get someone to do a seven-hour experiment where there's two blocks in
907
00:58:21,600 --> 00:58:22,600
it now?
908
00:58:22,600 --> 00:58:23,600
Yeah.
909
00:58:23,600 --> 00:58:27,520
So yeah, it's someday, hopefully I'll figure out a way to make it work, but that's my dream
910
00:58:27,520 --> 00:58:31,480
is to do these things and systematically manipulate attention.
911
00:58:31,480 --> 00:58:36,040
Yeah, but I think you'd have to do it in a way that didn't, that have to be a little bit
912
00:58:36,040 --> 00:58:37,920
orthogonal to the MMN itself, right?
913
00:58:37,920 --> 00:58:42,560
So, I don't think you could, if you probe them on the actual thing you care about, like that
914
00:58:42,560 --> 00:58:48,240
tense feature, I think that would really cause it to not even be in MMN generating situation.
915
00:58:48,240 --> 00:58:52,440
Like you probably would have to have them attend to a different aspect of the stimuli
916
00:58:52,440 --> 00:58:55,920
and throw away those trials and yeah, I'm guessing.
917
00:58:55,920 --> 00:59:01,480
Like every time it starts with an S or like one of these like P300 paradigms where they've
918
00:59:01,480 --> 00:59:04,240
got a standard deviant and a target.
919
00:59:04,240 --> 00:59:13,040
Yeah, so if we, leaving aside our quibbles about the timing and going with your conclusion
920
00:59:13,040 --> 00:59:19,280
that you've demonstrated MMNs for this very abstract contrast, what does that tell you
921
00:59:19,280 --> 00:59:22,160
about the MMN that we didn't know previously?
922
00:59:22,160 --> 00:59:26,640
How does this paper move the field forward?
923
00:59:26,640 --> 00:59:31,760
I mean, honestly, I'm not sure it tells us things we didn't know previously.
924
00:59:31,760 --> 00:59:37,160
The way I think of it is it tells us things that we had assumed but had not actually demonstrated.
925
00:59:37,160 --> 00:59:44,200
So, like if we had not gotten an MMN that would have been sad for me as a linguist, but
926
00:59:44,200 --> 00:59:48,640
it actually would have really changed what we know about the MMN because we've spent
927
00:59:48,640 --> 00:59:53,840
however, many decades thinking the MMN is sensitive to these abstract contrasts and that would
928
00:59:53,840 --> 00:59:56,600
have shown that it actually isn't.
929
00:59:56,600 --> 01:00:01,320
I think what this has shown us is that like the MMN is actually sensitive to the things
930
01:00:01,320 --> 01:00:07,520
that we've been saying it's sensitive to but haven't really proven and like now we've
931
01:00:07,520 --> 01:00:12,800
hopefully demonstrated that for at least one really abstract thing you really can get
932
01:00:12,800 --> 01:00:17,080
the MMN without needing like a physical correlate to it.
933
01:00:17,080 --> 01:00:24,760
So, it shows that the brain is like accessing these very abstract representations and generating
934
01:00:24,760 --> 01:00:32,280
categories based purely on abstract information even before you're paying attention to it.
935
01:00:32,280 --> 01:00:35,120
Yeah, great.
936
01:00:35,120 --> 01:00:36,120
Yeah, I think it's a step forward.
937
01:00:36,120 --> 01:00:42,000
I mean, I think there's a big difference between like actually demonstrating something in
938
01:00:42,000 --> 01:00:48,120
a really robust way versus just kind of thinking it because you had experimentally pointed
939
01:00:48,120 --> 01:00:53,720
in that direction that weren't really but had been open to other interpretations.
940
01:00:53,720 --> 01:00:57,400
And would you, what was it like when you went back with your data to the reviewers?
941
01:00:57,400 --> 01:01:01,600
Was it easy to get the final version published compared to how it might have been coming in
942
01:01:01,600 --> 01:01:02,600
fresh?
943
01:01:02,600 --> 01:01:10,800
I think one sort of possible issue was the fact that we could not fully follow our stage
944
01:01:10,800 --> 01:01:12,840
one.
945
01:01:12,840 --> 01:01:19,280
If we did follow, I think we ended up, if we follow everything to a P, including the exclusion
946
01:01:19,280 --> 01:01:26,000
criteria out of 70 something participants, I think we would end up with only 15 or 16
947
01:01:26,000 --> 01:01:28,960
because it was extreme restrict.
948
01:01:28,960 --> 01:01:39,720
So, we had to justify these changes and sort of mention why and also do separate sub analysis,
949
01:01:39,720 --> 01:01:47,160
including just these participants who would be included in our previously split criteria.
950
01:01:47,160 --> 01:01:51,200
But I think outside of that, I think it was quite straightforward.
951
01:01:51,200 --> 01:01:52,200
Don’t you think?
952
01:01:52,200 --> 01:01:59,520
Yeah, I think like, yeah, we had to have a little extra step of sort of get making actually
953
01:01:59,520 --> 01:02:02,400
even before we did the paper.
954
01:02:02,400 --> 01:02:06,040
I think we had checked with the editor and reviewers like, are you okay with us doing
955
01:02:06,040 --> 01:02:08,920
the analysis in this different way?
956
01:02:08,920 --> 01:02:14,280
And then, yeah, other than that, the second stage of review for a registered report is really
957
01:02:14,280 --> 01:02:19,200
painless because they're not judging, is it a good paper or anything?
958
01:02:19,200 --> 01:02:23,760
They're just judging, did you follow what you said you would follow and are your conclusions
959
01:02:23,760 --> 01:02:26,560
consistent with what you said they would be?
960
01:02:26,560 --> 01:02:31,600
And so, it was like, probably the most painless review process I've ever had.
961
01:02:31,600 --> 01:02:38,360
Thanks to all the previous work that had been put in at the original registration time.
962
01:02:38,360 --> 01:02:42,680
So, would you do it again, registered report?
963
01:02:42,680 --> 01:02:46,880
That's a good, we've talked, Bernard and I have talked about that a lot.
964
01:02:46,880 --> 01:02:51,080
So, there are a lot of benefits to it that we talked about.
965
01:02:51,080 --> 01:02:55,880
And we did, we have another one that I think is just about to come out.
966
01:02:55,880 --> 01:03:00,680
But we also felt like there are some situations, it's not for everything, right?
967
01:03:00,680 --> 01:03:03,400
Not every paper has to be a registered report.
968
01:03:03,400 --> 01:03:08,160
So, I remember Bernard, there were a few issues we had talked about for what were they
969
01:03:08,160 --> 01:03:11,440
like situations where it might actually not be the best thing?
970
01:03:11,440 --> 01:03:19,240
Yeah, I think for early career researchers, if you do stage one, it's a bit time consuming.
971
01:03:19,240 --> 01:03:25,080
For this paper, the stage one was admitted in 2021 and then we only got the paper out by
972
01:03:25,080 --> 01:03:26,880
mid-2024.
973
01:03:26,880 --> 01:03:32,840
So, it is quite a long process, though admittedly, a lot of that is us recruiting and testing
974
01:03:32,840 --> 01:03:39,920
participants, but still, it is, it takes much longer than a standard paper that you would
975
01:03:39,920 --> 01:03:41,200
do.
976
01:03:41,200 --> 01:03:49,240
And also, after you've done all this stage one work, which arguably could be more or might
977
01:03:49,240 --> 01:03:54,480
take more time than the stage two actually, you don't really have anything to show for
978
01:03:54,480 --> 01:03:58,240
it as a sort of early-stage researcher.
979
01:03:58,240 --> 01:04:02,400
So, you're just, you just have, sort of, end up with nothing.
980
01:04:02,400 --> 01:04:08,120
Although this is different from journal to journal, some journals do publish the stage one as registered
981
01:04:08,120 --> 01:04:10,880
report protocols.
982
01:04:10,880 --> 01:04:21,720
And also, I do feel as though for these registered reports, if it is something that is sort of well
983
01:04:21,720 --> 01:04:28,080
founded and quite straightforward, it might be worth just doing straight on without the stage
984
01:04:28,080 --> 01:04:29,080
one process.
985
01:04:29,080 --> 01:04:30,080
Yeah.
986
01:04:30,080 --> 01:04:35,320
And the other thing we had talked about was like, there, I think there's a lot of benefit
987
01:04:35,320 --> 01:04:40,640
to like register report for the kind of thing we did where we are like we have a really
988
01:04:40,640 --> 01:04:43,880
specific prediction and we want to see it.
989
01:04:43,880 --> 01:04:48,200
But sometime, like there's some research where you just need to be able to like chase the
990
01:04:48,200 --> 01:04:52,000
rabbit and not expecting where things are going to go.
991
01:04:52,000 --> 01:04:57,320
So, like my first MMN paper, which was with Kevin Schluter, like was kind of like that,
992
01:04:57,320 --> 01:05:02,000
we did an experiment trying to test one thing and this unrelated weird other thing happened
993
01:05:02,000 --> 01:05:03,920
and we were like, why did that happen? (Laughter)
994
01:05:03,920 --> 01:05:07,560
And so, we did three more experiments trying to figure out what happened.
995
01:05:07,560 --> 01:05:11,320
And it was like some of the most fun I've ever had doing science.
996
01:05:11,320 --> 01:05:15,240
And I like it couldn't really have been read pre-registered because it was like we were
997
01:05:15,240 --> 01:05:18,960
coming up with questions as each new set of data came out.
998
01:05:18,960 --> 01:05:23,880
So, I feel like it's got to be like a rich tapestry, like some research is really amenable
999
01:05:23,880 --> 01:05:29,160
to register reports, but there's also a lot of valuable to like doing exploration as things
1000
01:05:29,160 --> 01:05:30,480
are coming.
1001
01:05:30,480 --> 01:05:31,720
I definitely agree with that.
1002
01:05:31,720 --> 01:05:33,440
I think that we need both.
1003
01:05:33,440 --> 01:05:34,440
We need both.
1004
01:05:34,440 --> 01:05:37,400
Yeah, I'm definitely a rabbit chaser.
1005
01:05:37,400 --> 01:05:39,840
I don't, I rarely have hypotheses.
1006
01:05:39,840 --> 01:05:43,160
Like my whole life is just a fishing expedition.
1007
01:05:43,160 --> 01:05:45,440
And that's just the way I do science.
1008
01:05:45,440 --> 01:05:49,920
And like it definitely like, but rub some people the wrong way, but like, I don't know,
1009
01:05:49,920 --> 01:05:51,640
like there is room for all, right?
1010
01:05:51,640 --> 01:05:58,440
And I think especially when it comes to clinical applications, like I've started to be very
1011
01:05:58,440 --> 01:06:02,560
really feel strongly that like any clinical claims need to be pre-registered just because
1012
01:06:02,560 --> 01:06:06,200
the temptation to p-hack is just too great.
1013
01:06:06,200 --> 01:06:10,760
And if you know, if it's really, you know, going to apply to healthcare and you're making
1014
01:06:10,760 --> 01:06:15,680
healthcare decisions based on it, then I want to know that it was done in a really rigorous
1015
01:06:15,680 --> 01:06:16,680
way.
1016
01:06:16,680 --> 01:06:21,040
So, I think that there's, I think you make good points and there's really a whole landscape
1017
01:06:21,040 --> 01:06:22,800
of different kinds of studies.
1018
01:06:22,800 --> 01:06:30,400
Oh, I think one other thing that we wanted to share was also the fact that some editors and
1019
01:06:30,400 --> 01:06:37,720
reviewers might not be as familiar with the registered report process and on how to review
1020
01:06:37,720 --> 01:06:39,880
stage ones and stage two.
1021
01:06:39,880 --> 01:06:45,640
So, we had really good experience with Neurobiology of Language, but in other journals, this might
1022
01:06:45,640 --> 01:06:50,760
this might be sort of unfamiliar format and your stage two might be reviewed just like a traditional
1023
01:06:50,760 --> 01:06:51,760
paper.
1024
01:06:51,760 --> 01:06:57,120
So, after you report this, after your stage one is accepted, your stage two, you might still
1025
01:06:57,120 --> 01:07:03,680
get comments from reviewers on design, for example, which is probably not justified at that
1026
01:07:03,680 --> 01:07:04,680
stage.
1027
01:07:04,680 --> 01:07:12,000
So that is also another sort of challenge if you want to try your hand and do register reports.
1028
01:07:12,000 --> 01:07:14,480
Yeah, that would be very problematic.
1029
01:07:14,480 --> 01:07:16,800
I'm glad that you didn't have that experience here.
1030
01:07:16,800 --> 01:07:22,440
I mean, I think that because Neurobiology of Language is trying to be a more forward looking
1031
01:07:22,440 --> 01:07:27,520
journal, hopefully anybody that's going to be editing one of those would shut down any
1032
01:07:27,520 --> 01:07:32,000
reviewer that tried to go in that direction and you'd never even see that.
1033
01:07:32,000 --> 01:07:34,960
I would hope that you would never even see that review or comment, right?
1034
01:07:34,960 --> 01:07:40,200
It should be filtered out by the editor, but I'm glad it didn't go that way.
1035
01:07:40,200 --> 01:07:41,200
Yeah.
1036
01:07:41,200 --> 01:07:42,560
All right, great.
1037
01:07:42,560 --> 01:07:47,520
Well, thank you guys so much for taking the time to chat with me about this paper.
1038
01:07:47,520 --> 01:07:51,520
I really enjoyed reading it and I really enjoyed talking to you guys about it.
1039
01:07:51,520 --> 01:07:53,040
Thank you for having us.
1040
01:07:53,040 --> 01:07:54,040
Thanks for having us.
1041
01:07:54,040 --> 01:07:59,040
Yes, great experience to be on this and I listened to a lot of other great discussions
1042
01:07:59,040 --> 01:08:01,000
on this, so I'm glad we can be part of it.
1043
01:08:01,000 --> 01:08:02,000
Oh, thank you.
1044
01:08:02,000 --> 01:08:03,000
I'm glad you've enjoyed it.
1045
01:08:03,000 --> 01:08:04,000
All right.
1046
01:08:04,000 --> 01:08:05,000
Have a good rest of your day.
1047
01:08:05,000 --> 01:08:06,000
I'll see you guys later.
1048
01:08:06,000 --> 01:08:07,000
Thanks.
1049
01:08:07,000 --> 01:08:08,000
We'll see you.
1050
01:08:08,000 --> 01:08:09,000
Have a good one.
1051
01:08:09,000 --> 01:08:10,000
Bye-bye.
1052
01:08:10,000 --> 01:08:11,000
Bye.
1053
01:08:11,000 --> 01:08:13,000
That's it for episode 33.
1054
01:08:13,000 --> 01:08:16,720
Thanks very much, Steve and Bernard for joining me from Kansas and Hong Kong.
1055
01:08:16,720 --> 01:08:17,720
Very much appreciated.
1056
01:08:17,720 --> 01:08:21,600
I've linked to the paper we discussed in the show notes and on the podcast website at
1057
01:08:21,600 --> 01:08:24,560
langneurosci.org/podcast
1058
01:08:24,560 --> 01:08:28,600
Thanks, as always to Marcia Petyt for her tireless work on transcribing the podcast, which I really
1059
01:08:28,600 --> 01:08:32,240
appreciate as it brings out discussions to a wider audience.
1060
01:08:32,240 --> 01:08:33,480
Bye for now and see you next time.