Name Conditioning in Event-Related Brain Potentials

Four experiments are reported in which two harmonic tones (CS+ and CS-) were paired with a participant’s own name (SON) and different names (DN), respectively. A third tone was not paired with any other stimulus and served as a standard (frequent stimulus) in a three-stimuli oddball paradigm. The larger posterior positivity (P3) to SON than DN, found in previous studies, was replicated in all experiments. Conditioning of the P3 response was albeit observed in two similar experiments (1 and 3), but the obtained effects were weak and not identical in the two experiments. Only Experiment 4, where the number of CS/UCS pairings and the Stimulus-Onset Asynchrony between CS and UCS were increased, showed clear CS+/CS- differences both in time and time-frequency domains. Surprisingly, differential responses to CS+ and CS- were also obtained in Experiment 2, although SON and DN in that experiment were masked and never consciously recognized as meaning words (recognition rate 0/63 participants). The results are discussed in the context of other ERP conditioning experiments and, particularly, the studies of non-conscious effect on ERP. Several further experiments are suggested to replicate and extend the present findings and to remove the remaining methodological limitations.

patients with severe brain damage, in the present study we intended to explore the effect of classical 48 conditioning on ERP to simple stimuli. Relatively simple harmonic tones were chosen as CS, based on the 49 finding that harmonic tones elicit more distinct and stable ERP effects than sine tones in both healthy 50 individuals (Tervaniemi et al., 2000) and neurological patients (Kotchoubey et al., 2003). An individual's 51 own name, which has been suggested to possess particular significance for the individual, was used as a 52 non-aversive UCS. The effects of a subject's own name (SON) on ERPs have been established in normal 53 populations (Fischler et al., 1987) and severely brain-injured patients (Perrin et al., 2006), in waking state 54 (Holeckova et al., 2006) and during sleep (Perrin, Garcia-Larrea, Mauguiere, & Bastuji, 1999). In a three-55 stimulus oddball, in which SON and a control stimulus (usually, a different name: DN) are presented as 56 two rare stimuli, SON elicits a larger P3 component than DN (e.g., Perrin et al., 2006;Kotchoubey et al., 57 2004). We expected to obtain a similar effect in response to harmonic tones (presented as CS) paired 58 with names. 59 Methods: General 60 4 Three different groups of healthy participants took part in the study: one group (nine males and 14 61 females, aged 22-29) in Experiment 1, the second group (nine males and 13 females, aged 22-29) in 62 Experiments 2 and 3, and the third group (twelve males and 13 females, aged 19-42) in Experiment 4. In 63 the second group Experiment 2 always preceded Experiment 3. Data of two males in Experiment 2 were 64 excluded (thus the group contained 7 males). 65 None of the participants had had any disease of the nervous system or hearing disorders in the past, or 66 reported use of any drugs during the last week before the experiment. Participants were seated in a 67 comfortable chair and asked to close their eyes and to listen attentively to the stimuli. Informed consent 68 was obtained from each participant. The study was approved by the Ethical Committee of the University 69 of Tübingen. 70 The EEG in all experiments was recorded using 64 active ActiCHamp electrodes (Easycap GmbH, 71 Herrsching, Germany) located according to the extended 10-20 system. The vertical and horizontal 72 electrooculagram were also recorded. The resistance was below 15 kOhm. Online reference was at Cz, 73 offline re-referenced to average mastoids. The digitalization rate was 1000 Hz. 74 Off-line inspection of the recordings revealed in some traces poor data quality in one or two of the 64 75 channels. These channels were replaced with interpolation of the adjacent electrodes. After this, an 76 Independent Component Analysis (ICA) was employed for each participant to separate and remove 77 activity due to ocular artifacts using the AMICA algorithm (Palmer et al., 2012). Components clearly 78 related to eye movements were removed using EEGLAB. Additionally, components that were mapped 79 onto one electrode and could be clearly distinguished from EEG signals were subtracted from the data. 80 EEG segments that still contained artifacts after ICA correction were dismissed. The ERPs were filtered 81 within a band from 0.1 to 30 Hz and averaged in relation to a baseline from -200 ms to 0 ms. As we 82 supposed that the responses would change during the roughly 7-minute test phase, ERPs were averaged 83 separately for the first, second, and third thirds of the whole sequence of 400 stimuli . The three periods,  84   corresponding to stimuli 1 -133, stimuli 134 -266, and stimuli 267 -400, will be referred to as T1, T2,  85 and T3, respectively. Each average included at least 18 (usually 20) CS+ and CS-. 86 The amplitudes of ERP components were measured as the area under the curve within the time windows For brevity, the present report describes only those data that are related to the critical comparison 97 between the CS+ and CS-responses at the midline electrodes Fz, Cz, and Pz, were the effects were best 98 pronounced. The statistical analysis was performed using a repeated measures ANOVA with factors 99 Stimulus, Site, and Time. When appropriate, we used Greenhouse-Geisser non-sphericity correction for 100 degrees of freedom. 101

Methods 103
ERPs were recorded to three chords, each consisting of five harmonic tonal frequencies (e.g., 330, 660, 104 1320, 2640, and 5280 Hz). One of the chords was used as standard and the other two served as CS+ and 105 6 times with three different names (DN). All names were spoken with the official German pronunciation by 107 a female speaker, not familiar to any participant. The control names originated from the same pool of 108 the most frequent German names used for each subject's own name. They always had a very similar 109 duration as the own name (means 669 ms and 676 ms) and contained the same number of syllables. The 110 standard was presented 21 times, not accompanied by any other stimulus. Tone duration was 200 ms, 111 and the intensity was 75 dB above the average threshold. The stimulus-onset-asynchrony (SOA) within a 112 pair tone-name was 300 ms. The SOA after a tone-word pair was 1700-1800 ms, and after standards it 113 was 1150-1250 ms. All stimuli were presented binaurally through aerodynamic earphones, in a 114 pseudorandomized order, in which none of the three tones appeared more than three times in a row. 115 In the test phase, which followed immediately after the acquisition phase, the standard was presented 116 280 times, and CS+ and CS-60 times each. No other stimuli were presented. The SOA varied between 117 950 and 1050 ms. The order of presentation was randomized except that CS+ and CS-could not be 118 delivered more than twice in a row. 119

Results 120
All participants reported after the experiment, that they had heard "two or three" different harmonic 121 tones, and that at the beginning one of the tones was linked to their own name, and another tone, to 122 other names. 123 The acquisition phase replicated the already known effect of a larger P3 to SON. The effect was most 124 clear in the P3a window (F(1,22) = 11.55, p = .003,  2 = .34). Also in the LTW, the amplitude was negative 125 to DN but positive to SON at Cz and Pz, yielding a significant Stimulus x Site interaction: F(2,44) = 5.37, p 126 = .014,  2 = .20. Importantly, the P3a amplitude was larger to CS+ than CS-(mean amplitudes 3.47 versus 127 1.78 µV; F(1,22) = 10.63, p = .004,  2 = .33). The P3(a) effect, was, however, instable and disappeared in 128 7 the test phase, in which no differences between CS+ and CS-responses were observed. The average ERP 129 waveforms are shown in Figures 1 and 2. 130

Experiment 2 131
Methods 132 In Experiment 2 the names were completely masked while preserving their acoustical features. The first 133 25% of time points of an original name were multiplied by a linearly spaced vector of coefficients from 134 1.5 to 0, and the remaining 75% points were set to 0. Then, the first 25% of time points of the same 135 name played backwards were multiplied by a linearly spaced vector of coefficients from 0 to 1.5, and the 136 last 75% time points remained unchanged. Finally, the two files were added. This technique permitted to 137 attain the same intensity-by-time dynamics as in the original names. In a pilot experiment the stimuli 138 were presented to forty healthy participants. None of them was able to recognize any name including 139 their own. 140

Methods 151
Although CS+ and CS-differed in the perceived pitch, in Experiment 3 they also were delivered 152 monaurally in two different ears to further increase their discriminability. The side of presentation was 153 counterbalanced among the participants. 154

Experiment 4 164
Methods 165 This Experiment entailed only one phase, during which the standard was presented 280 times, and CS+ 166 and CS-60 times each. Tone duration was 100 ms including 5 ms rise/fall phase. All CS were followed by 167 the corresponding UCS with SOA of 600 ms. This design aimed at the recording of the late ERP 168 components to CS immediately during acquisition, because in the other experiments (with SOA of 300 169 ms) they could be measured only in the test phase. presented 20 times each), the same difference is also significant for P3a: F(1,24) = 5.36, p = .03,  2 = .18. 180

Time-frequency analysis 181
As can be seen in Figure 3 To summarize, these findings indicate that classical conditioning of electrophysiological responses using 202 the own name as UCS is possible. The differential responses to CS+ and CS-in Experiment 4 were 203 demonstrated in both time (i.e., late ERP positivities) and time-frequency domains (i.e., higher theta but 204 lower alpha activity to CS+ than CS-). Importantly, the better differentiation between the two CS was 205 found in the same experiment in which the two UCS were also best distinguished. Generally, however, 206 learning effects were relatively weak and unstable. Experiment 4, in which the clearest learning effects 207 were obtained, did not contain extinction. The only effect that survived extinction was the differential 208 response in the LTW in Experiment 3. Therefore, we do not believe that the paradigm can be used in 209 clinical applications in its present form; rather, further methodological work is needed. Another 210 limitation concerns the theoretical interpretation. Although the effect was primarily concentrated 211 around P3a, we cannot rule out that other ERP components (e.g., P2 or P3b) also contributed to its 212 generation. Also this issue should be followed in further experiments. 213 On the background of these largely predicted findings, the results of Experiment 2 were fully 214 unexpected. Remember the masked names from this experiment were presented to a total of 63 215 individuals (40 participants of the pilot experiment + 22 participants of Experiment 2 + the first author 216 who was unaware about masking developed by the second author), and none of them (0/63) recognized 217 any name including their own. The stimuli were not even perceived as meaningful words. ERP, however, 218 demonstrated a significantly larger positivity to SON than DN. Notably, this differential response was 219 about 100 ms later than in all three experiments where SON was consciously recognized, indicating the 220 presence of some additional processing operation(s). Although conditional responses in Experiment 2 did 221 not differ during acquisition, the test phase revealed a significant P3a effect, quite similar to that in the 222 acquisition phase of Experiment 1 and 4, and additionally, a very strong ( 2 = .50) enhancement of the 223 N1 amplitude to CS+ as compared with CS-. This N1 effect was very consistent at the individual level, 224 being observed in 20 of the 22 participants. 225 During acquisition, SON was presented as UCS+ with the same frequency as three DN as UCS-. Therefore, 226 each individual DN was presented three times less frequently than SON. Although this fact is a 227 methodological limitation of the present study, it can hardly be responsible for any expected or 228 surprising result. Firstly, this arrangement was the same in all four experiments, but the results were 229 different. Secondly, N1 and P3a are expected to be larger to rare than frequent stimuli. If N1 is 230 superposed by a Mismatch Negativity, this wave is also larger to rare than frequent stimuli. On this basis 231 we might expect larger amplitudes to CS-(previously linked to rare DN) than to CS+ (previously linked to 232 the more frequent SON), but the opposite was found. Only the non-significant increase of P3a in 233 Experiment 3 would be in line with the frequentist interpretation. But even in this case such 234 interpretation meets a considerable problem: it is fully unclear why ERP responses differ between the 235 two CS having equal frequencies (and only linked to stimuli of different frequencies), whereas these 236 responses do not differ between the two UCS having different frequencies. 237 If explanations related to experimental methodology are rejected, the only possible interpretation of the 238 data of Experiment 2 remains that participants subconsciously distinguished between SON and DN even 239 though they did not recognize them. This is particularly possible because our masking technique assured 240 the similar intensity/time function of unmasked and masked UCS. If, for example, a sound file consisted 12 of 100 data points, the first point was the same in masked and unmasked stimuli, the second point 242 contained 95.8% common information, the third point 91.7%, etc. It may, therefore, be speculated that 243 participants unconsciously recognized different personal significance of the stimuli even though they did 244 not identify their content. 245 The fact that stimuli that are not consciously recognized can nonetheless elicit significant ERP effects has 246 been shown in numerous studies (reviews Shevrin, 2001;Dehaene et al., 2006). Positivity is plotted upwards. 341