Contingency awareness, aging, and the parietal lobe

Abstract

Contingency awareness is thought to rely on an intact medial temporal lobe and also appears to be a function of age, as older subjects tend to be less aware. The current investigation used functional magnetic resonance imaging, transcranial direct current stimulation, and eyeblink classical conditioning to study brain processes related to contingency awareness as a function of age. Older adults were significantly less aware of the relationship between the tone-airpuff pairings than younger adults. Greater right parietal functional magnetic resonance imaging activation was associated with higher levels of contingency awareness for younger and older subjects. Cathodal transcranial direct current stimulation over the right parietal lobe led to lower levels of awareness in younger subjects without disrupting conditioned responses. Older adults exhibited hyperactivations in the parietal and medial temporal lobes, despite showing no conditioning deficits. These findings strongly support the idea that the parietal cortex serves as a substrate for contingency awareness and that age-related disruption of this region is sufficient to impair awareness, which may be a manifestation of some form of naturally occurring age-related neglect.

1. Introduction

Conscious awareness has been extensively discussed by scholars ranging from philosophers to scientists. Modern neuroscience continues to address the biological nature of human awareness, but the neuroanatomy supporting general awareness can be difficult to identify because awareness itself is a broad construct and consists of multiple psychological processes. However, one form of awareness, contingency awareness, specifically refers to the explicit knowledge of the temporal relationship between 2 events (Lovibond and Shanks, 2002).

One simple and elegant paradigm that has been used to investigate contingency awareness is classical conditioning. It is a well-characterized model system, in which a neutral conditioned stimulus (CS) and a biologically meaningful unconditioned stimulus (US) are temporally paired and has highly predictable behavioral outcomes. After multiple paired CS–US presentations, the CS alone elicits a conditioned response (CR) in anticipation of the US, indiacating that an association between the CS and US has been formed.Delay and trace conditioning are 2 different procedures that vary in timing. In delay conditioning, the CS and US coterminate, whereas in trace conditioning, a stimulus-free period (called the trace interval) elapses between offset of the CS and onset of the US. Interestingly, this minor manipulation has strong implications on what neural mechanisms are recruited. Human functional magnetic resonance imaging (fMRI) studies have shown that successful trace conditioning uniquely engages supplementary neural structures such as the medial temporal lobes (MTLs), hippocampus, parietal lobe, and middle frontal gyrus (Buchel et al., 1999; Cheng et al., 2008; Haritha et al., 2013; Knight et al., 2004).

One form of classical conditioning, eyeblink classical conditioning, is one of the most studied forms of mammalian learning, and consequently, the neuroscience community largely agrees that the cerebellum is critically important (Christian and Thompson, 2003). However, there is less agreement on the role of contingency awareness during human eyeblink conditioning as it has been extensively debated (LaBar and Disterhoft, 1998; Lovibond and Shanks, 2002; Manns et al., 2002). Clark and Squire (1998) argued that the hippocampus and awareness are necessary for trace conditioning, as temporal lobe amnesics, who could not accurately and explicitly report CS-US relationships, were impaired at trace but not delay conditioning. In addition, only healthy subjects classified as aware were able to demonstrate trace conditioning, whereas both aware and unaware subjects were capable of showing intact delay conditioning. This general claim that trace, but not delay, conditioning requires contingency awareness has been supported by subsequent studies using several variants of these procedures (Manns et al., 2001, 2000a,b, 2002; Smith et al., 2005).

Others have reported contradictory findings and argued that awareness is necessary for both trace and delay conditioning. Lovibond et al have published several studies that used the same methodology as Clark and Squire (1998) (Lovibond et al., 2011), masking procedures (Weidemann et al., 2013), the Perruchet effect (Weidemann and Lovibond, 2016), and verbal instructions (Weidemann et al., 2016) to manipulate awareness and all conclude that awareness was needed for both trace and delay conditioning. Furthermore, in a series of experiments examining the effects of aging and awareness on eyeblink conditioning, Knuttinen et al. (2001) and Bellebaum and Daum (2004) support the claim that awareness plays an important role for delay conditioning. Functional MRI studies using fear conditioning as a model showed that awareness of the CS-US relationship is associated with activations in the hippocampus, parahippocampus, middle frontal gyrus, and ventral striatum (Baeuchl et al., 2019; Cacciaglia et al., 2015; Carter et al., 2006; Klucken et al., 2009; Knight et al., 2009; Tabbert et al., 2006, 2011).

The debate over the role of awareness during classical conditioning may be related to why some subjects fail to achieve contingency awareness. There may be several reasons for why contingency awareness eluded some older subjects. One possible explanation is that the ability to form accurate judgments for temporal relationships between 2 stimuli worsens as one ages (Bedard and Barnett-Cowan, 2016; Poliakoff et al., 2006). Importantly, this impairment was not due to a general sensory processing deficit as older adults performed comparably to younger adults during short time intervals (70 ms) (Setti et al., 2011). Another factor potentially contributing to a lack of awareness is that bottom-up forms of spatial attention necessary for conscious perception/awareness rely on fronto-parietal networks, which have been shown to be damaged in older populations (Chica and Bartolomeo, 2012; Hoffman and Morcom, 2018). Furthermore, it is also possible the CS and US are superficially processed and to some degree, neglected, and hence, this information fails to reach consciousness. Contingency awareness also appears to be a function of age, as older subjects tend to be less aware (Knuttinen et al., 2001) and have poorer trace conditioning relative to younger subjects (Finkbiner and Woodruff-Pak, 1991). To date, no fMRI studies have addressed the interaction between aging and contingency awareness using eyeblink conditioning as a model system. In 2 separate studies, we used fMRI and transcranial direct current stimulation (tDCS) to examine key brain structures, including the parietal lobe, a crucial area in neglect patients (Donaldson et al., 2015; Vuilleumier, 2013), as they relate to awareness in older and younger adults. If a lack of contingency awareness is partially mediated by neglect of the CS and US, greater parietal activation should be measured in those who were aware relative to unaware subjects. Because both awareness and the hippocampus are typically relied on during trace, but not delay conditioning, we predict that MTL activation would be greater for subjects who were aware relative to unaware subjects. Furthermore, because older subjects tend to demonstrate poorer awareness and trace conditioning in comparison to younger subjects, we predict that older subjects would be less aware than younger subjects and awareness would be correlated with levels of trace, but not delay conditioning. Our results provide partial support for these predictions and reveal that the neural substrates of CS-US contingency awareness likely involve the parietal region.

2. Materials and methods

2.1. Subjects

Subjects were recruited from the greater Baltimore area through the internet (Craigslist), flyers, and local TV advertisements. 49 healthy, right-handed volunteers (19 males) participated in the fMRI study: 28 younger adults (24.6 ± 0.6 years) and 21 older adults (63.7 ± 0.8 years). 16 healthy, right-handed volunteers (5 males; 24.3 ± 0.7 years) participated in the tDCS study. Selection was based on the following exclusion criteria: (1) disturbed consciousness; (2) other neurological or systemic disorders that can cause dementia or cognitive dysfunction; (3) prior history of a major psychiatric disorder; (4) history of definite stroke; (5) focal lesion on MRI examination; (6) use of anxiolytic, antidepressant, neuroleptic, or sedative medication. General cognitive abilities were assessed with the Mini-Mental State Examination (Folstein et al., 1975); a simple reaction time task; the Alzheimer’s Disease Assessment Scale (ADAS: Recall and Recognition) (Rosen et al., 1984); subtests of the Wechsler Adult Intelligence Scale–III (Matrix Reasoning, Digit Span, Digit Symbol, and Vocabulary) (Wechsler, 1987); and the National Adult Reading Test (Nelson, 1982). Scores from the Wechsler Adult Intelligence Scale–III were age-adjusted for each participant’s age. Each subject received an audiogram at 3 frequencies (500, 1000, and 1500 Hz) in each ear to determine their minimum auditory threshold. All subjects were compensated $20/h, and all procedures were approved by the institutional review boards for human subject research at the Johns Hopkins University School of Medicine.

2.2. Magnetic resonance imaging

Whole-brain imaging was performed on a Philips 3-tesla MRI scanner. Functional images were collected using a T2*-weighted gradient echo planar imaging pulse sequence. 6-mm axial slices (TR, 1000 ms; TE, 30 ms; FOV, 24 cm; flip angle, 61°) were collected in a series of 1090 sequential images. Structural images were collected using a T1-weighted magnetization-prepared rapid acquisition gradient echo pulse sequence.

2.3. Transcranial direct current stimulation

tDCS (2 mA) was delivered with 0.9% saline-soaked sponges (25 cm²) over the right parietal lobe on the participant’s scalp with the reference electrode on the left mandible. Cathodal stimulation lasted 20 minutes and sham stimulation lasted 30 seconds. Localization was based on peak activation coordinates (x, y, z) in the fMRI study. Scalp location over this coordinate was determined on one participant whose structural MRI was coregistered to head land-marks using Brainsight (Rogue Industries, Canada). For normalization purposes, this location was defined as a percentage of the anterior-posterior (nasion to inion) and left-right (preauricular points) axes. The normalized scalp location (39.5% of the distance between the nasion-inion as measured from the inion and 22.6% of the distance between the left-right preauricular points as measured from the right preauricular point) was then applied to individual subjects.

2.4. Eyeblink conditioning

A laptop computer interfaced to DT9834 data acquisition module (Data Translations) running custom software developed under LabView version 7.1 (National Instruments) was used for presenting stimuli (tones and airpuffs) and recording eyeblinks. Auditory stimuli were delivered with MRI-compatible pneumatic head-phones (MRA, Inc). Custom modifications to standard laboratory safety glasses accommodated the end of a polyethylene tube for airpuff delivery and an MRI-compatible infrared sensor for recording eyeblinks (Cheng et al., 2008). A fiber-optic probe (RoMack, Inc) measured the reflectance of infrared light from the left eye (Miller et al., 2005), and airpuff delivery was controlled by a solenoid valve (Asco).

2.5. Stimuli

A 1000-Hz tone served as the delay and trace CS. The delay CS lasted 1350 ms and coterminated with a 100 ms left corneal airpuff (5 psi). The trace CS lasted 250 ms and was followed by a 1000 ms trace period (stimulus free) before airpuff presentation (Fig. 1A). For the MRI study, 120 delay or trace CS-US presentations were delivered with an average intertrial interval of 18 seconds (range of 15–21 seconds). For the tDCS study, 30 delay CS-US presentations were delivered with an average intertrial interval of 18 seconds (range of 15–21 seconds). CS and US were paired at 100% for both studies.

Fig. 1.

Experimental design and behavioral data. (A) Participants received either delay CSs (coterminates with the US) or trace CSs (stimulus-free period between CS offset and US onset). Dotted lines indicate the time window (500 ms) in which eyeblinks were considered conditioned responses. (B) Learning during the early stages of the experiment. (C) No significant differences in overall conditioning levels (120 trials) were measured between the 4 groups of participants. (D) The left panel shows that older adults were less likely to be aware of the CS-US relationship as compared with younger adults. The right panel shows that older adults scored significantly lower than younger adults on their postexperimental questionnaire. Abbreviations: CR, conditioned response; CS, conditioned stimulus; US, unconditioned stimulus.

2.6. Procedures

2.6.1. MRI study

Subjects were randomly assigned to receive delay or trace conditioning trials. They were informed that this study investigated the effects of distracting tones and airpuffs have on their ability to remember details about a silent movie (Charlie Chaplin’s The Gold Rush). They were told that their movie knowledge would be tested after the experiment. After being fitted with the safety glasses, they were placed in the magnet with instructions to watch and pay attention to the movie while distracting tones and airpuffs were presented. After conditioning, a movie quiz and postexperimental questionnaire assessing awareness of the CS-US contingencies were administered (see Appendix 1).

2.6.2. tDCS study

Subjects were randomly assigned to receive cathodal or sham stimulation. They were informed that this study investigated the effects of brain stimulation, distracting tones, and airpuffs have on their ability to remember details about the silent movie. A line bisection test was administered before stimulation. Once scalp localization was determined on each participant, sponges and safety glasses were applied. Stimulation began simultaneously with the start of the movie and conditioning. After stimulation, a movie quiz, postexperimental questionnaire assessing awareness of the CS-US contingencies (see Appendix 1), and a second-line bisection test was administered.

2.7. Analyses

2.7.1. Behavioral data

CRs were defined using the following criteria: the difference between the maximum and minimum responses in a 500 ms pre-US time window must exceed 4 times the standard deviation of the mean of the baseline period (250 ms pre-CS presentation). The 500 ms pre-US time window was selected to minimize the inclusion of voluntary and alpha responses as CRs (Spence and Ross, 1959). Performance was expressed as %CR and examined in an age (younger and older) by trial type (delay and trace) analysis of variance (ANOVA). To evaluate early learning, the frequency at which subjects produced a CR during the first 4 trials was evaluated using a nonparametric Cochran’s Q test.

2.7.2. Awareness data

A postexperimental questionnaire consisting of 7 true/false statements (Manns et al., 2000a) was given to subjects to probe their awareness of the CS-US relationship. Subjects were classified as aware if they answered 6 or more questions correctly and unaware if they answered fewer than 6 questions correctly. This threshold was based on the probability of obtaining 6 correct responses of 7 from a binomial distribution that is set to p = 0.05. Chi-square analyses were performed on awareness level (aware or unaware) in younger and older subjects.

2.7.3. Imaging data

Statistical Parametric Mapping (SPM2 and SPM8) software (Wellcome Department of Cognitive Neurology, London, UK) was used to perform structural and functional imaging preprocessing and statistical analyses. Preprocessing included motion correction, slice timing correction, structural data coregistration, normalization, and smoothing. Echo planar imaging functional images were realigned and resliced correcting for minor motion artifacts, and structural images were coregistered to the mean motion-corrected functional image for each participant. Functional and structural images were transformed into standard stereotaxic space (2 × 2 × 2) according to the Montreal Neurological Institute protocol, and the functional images were smoothed with a Gaussian filter (full-width half-maximum 5 mm). Cerebellar data were treated separately by isolation and normalization (1 × 1 × 1) into standard stereotaxic space using the spatially unbiased atlas template of the human cerebellum and brainstem (Diedrichsen, 2006). First-level analyses adopted an event-related approach in which the general linear model was used to estimate individual subject activations based on all 120 trials. Reference waveforms were created based on CS onset times and were convolved with individually estimated hemodynamic response functions to account for aging effects on the hemodynamic response. These hemodynamic response functions were generated using a standard finger tapping task.

In a whole brain, voxel-wise analysis, between-group contrasts were performed to assess the effects of conditioning (delay and trace), awareness (aware vs. unaware), and aging (older vs. younger) (Tables 2–4). Additional whole-brain, voxel-wise analyses included contrasts between aware and unaware participants and regression analyses using individual awareness scores as predictors of interest. Second, structural regions of interest (ROI) analyses based on probabilistic maps of MTL structures (Amunts et al., 2005) and a priori ROI analyses using parietal lobe coordinates that overlapped in lesions in neglect patients (Mort et al., 2003) were performed. These coordinates served as the center of a 3D sphere (10 mm radius) that was created to individually sample the parietal region most affected in neglect patients. For voxelwise analyses, a p < 0.001 significant threshold with a minimum cluster threshold of 10 voxels was set. For each ROI analysis, mean beta values were extracted from individual participants so that statistical analyses could be performed using SPSS software. Finally, a correlation analysis between participant’s awareness scores and their parietal activity was performed.

Table 2.

Significant BOLD changes related to delay and trace conditioning

Brain structure (neocortex, p < 0.00001)	x	y	z	SPM{Z}	Voxels
L Superior temporal gyrus (BA 41, 22)	−55	−20	5	8	5135
R Superior temporal gyrus (BA 13, 41, 22)	54	−27	6	8	5976
R insula (BA 13)	38	−26	4	8
L precentral gyrus (BA 6)	−39	−11	40	4.8	110
R precentral gyrus (BA 4)	41	−15	41	5.6	232
R postcentral gyrus (BA 2)	32	−22	33	5.5	44
R cingulate gyrus (BA 24)	2	7	32	5.7	773
R medial frontal gyrus (BA 6)	2	3	50	5.4
R insula (BA 13)	31	20	7	6.2	354
L insula (BA 13)	−32	21	6	6.3	332
R lingual gyrus (BA 18)	19	−56	3	6.2	1853
L lingual gyrus (BA 18)	−11	−69	5	6.1
R thalamus	8	−26	−3	5.7
R cuneus (BA 18)	13	−77	22	5.9
L posterior cingulate (BA 30)	−20	−66	7	5.7
Brain structure (cerebellum, p < 0.001)	x	y	z	SPM{Z}	Voxels
L cerebellum (lobule HVI)	−34	−56	−32	3.2	11
L cerebellum (lobule HVIII)	−28	−66	−52	4.3	67
L cerebellum (lobule HVI)	−26	−62	−26	3.4	19
L cerebellum (lobule HVIII)	−18	−68	−40	3.3	14
R cerebellum (lobule HVI)	8	−68	−20	4	57
R cerebellum (lobule HVIII)	20	−70	−54	3.9	69

MNI (cerebellar) and Talairach (noncerebellar) coordinates of activation maxima (Schmahmann et al., 2000; Talairach and Tournoux, 1988) as a function of delay and trace trials.

Key: BOLD, blood oxygenation level dependent; MNI, Montreal Neurological Institute.

Table 4.

Significant BOLD differences between older and younger adults

Older > younger
Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
L superior temporal gyrus (BA 38)	−28	20	−30	4.4	73
L precuneus (BA 7)	−13	−64	30	3.5	25
L cingulate gyrus (BA 23)	1	−17	24	4	30
R cingulate gyrus (BA 24)	4	4	25	3.6	33
R parahippocampal gyrus (BA 35)	16	−25	−9	3.3	29
R parahippocampal gyrus (BA 34)	16	−6	−16	3.8	49
R middle frontal gyrus (BA 9)	32	40	36	3.5	20
R middle frontal gyrus (BA 46)	40	33	13	3.8	104
R superior parietal lobule (BA 7)	43	−61	50	3.6	37
R superior temporal gyrus (BA 38)	42	6	−20	3.6	21
R middle temporal gyrus (BA 21)	62	−21	−9	3.9	118
Younger > older
Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
None

Talairach coordinates of activation maxima (Talairach and Tournoux, 1988) as a function of age.

Key: BOLD, blood oxygenation level dependent; EBC, eyeblink classical conditioning.

3. Results

3.1. Behavioral results

To investigate awareness and possible instances of neglect during eyeblink classical conditioning, younger and older adults were randomly assigned to receive delay or trace conditioning during fMRI (Fig. 1A). Subject breakdown was 14 younger delay (24.4 ± 0.9 years), 14 younger trace (24.9 ± 0.8 years), 11 older delay (64.6 ± 0.8 years), and 10 older trace (62.7 ± 0.9 years). Eyeblinks were considered CRs if they occurred 500 ms before presentation of the corneal airpuff to exclude alpha, voluntary responses (Spence and Ross, 1959). Postexperimental questionnaires were administered after training, and each subject was labeled aware or unaware based on their answers.

Significant age-related differences in cognitive testing were seen in ADAS Recall (younger: 2.07 ± 0.29, older: 3.05 ± 0.27) Matrix Reasoning (younger: 13.32 ± 0.49, older: 11.43 ± 0.79), Digit Symbol (younger: 12.25 ± 0.69, older: 9.60 ± 0.77), and National Adult Reading Test (younger: 32.54 ± 1.11, older: 38.15 ± 1.63) (all p’s < 0.05). No significant changes were seen in Mini-Mental State Examination, reaction time, Digit Span, and Vocabulary (all p’s > 0.05). A trend (p = 0.051) was detected for ADAS recognition (younger: 0.82 ± 0.17, older 1.55 ± 0.35). See Table 1.

Table 1.

Cognitive testing results for younger and older adults

Cognitive testing	Younger (n = 28)	Older (n = 21)	t
Mini-Mental State Exam (MMSE)	29.36 (0.14)	29.43 (0.16)	0.34
Reaction time	281.19 (4.78)	292.66 (6.68)	1.44
Alzheimer’s Disease Assessment
Scale (ADAS)
Recall	2.07 (0.29)	3.05 (0.27)	2.38a
Recognition	0.82 (0.17)	1.55 (0.35)	2.00
Wechsler Adult Intelligence
Scale-III (WAIS-III)
Matrix Reasoning	13.32 (0.49)	11.43 (0.79)	2.14a
Digit Span	11.14 (0.46)	12.00 (0.67)	1.09
Digit Symbol	12.25 (0.69)	9.60 (0.77)	2.51a
Vocabulary	12.81 (0.48)	12.48 (0.36)	0.52
National Adult Reading Test (NART)	32.54 (1.11)	38.15 (1.63)	2.92a

^ap < 0.05 Values are mean (SEM).

A Cochran’s Q test indicated a significant difference between the number of subjects producing a CR during early trials (trials 1–4) (χ² (3, N = 49) = 4.98, p = 0.046). Pairwise comparisons showed that significantly more subjects produced a CR on trials 2, 3, and 4 relative to trial 1 (p ≤ 0.05) (Fig. 1B). However, when we assessed CR incidence (%CRs) on overall conditioning levels (120 trials), there were no significant main effects of age (F(1,41) = 0.418, p = 0.52), trial type (F(1,41) = 0.008, p = 0.93), and awareness (F(1,41) = 0.018, p = 0.90), or interactions (all p’s > 0.70) (Fig. 1C). A signicant association between age and awareness was calculated, χ² (1, N = 49) 4.98, p < 0.05. Despite comparable CR production, older adults were less likely to be classified as aware as younger adults (left Fig. 1D). Consistent with this classification measure, older adults also scored significantly (t(47) = 2.36, p = 0.02) lower than younger adults when raw awareness scores were compared (right Fig. 1D). Average auditory CS thresholds, unconditioned responses amplitudes, and number of correctly answered movie questions were compared and showed no significant differences between age groups (all p’s > 0.05). The behavioral difference in awareness scores guided subsequent imaging analyses.

3.2. Imaging results

3.2.1. Voxelwise analyses

3.2.1.1. Conditioning effects.

Conditioning effects, defined as significant whole-brain activations as a function of both delay and trace conditioning trial presentations, are summarized in Table 2. Large clusters of activation found in bilateral superior temporal gyri/auditory cortices (left: −55, −20, 5; right: 54, −27, 6) indicate that these regions play a role in processing the delay or trace tone CS, whereas greater activation in the right postcentral gyrus (32, −22, 33) may represent sensory information from the corneal airpuff as the left eye was trained. Greater activation in the bilateral precentral gyri (left: −39, −11, 40; right: 41, −15, 41) suggests these regions may reflect bilateral motor eyeblink responses (Campolattaro and Freeman, 2009), but we did not measure responses from the untrained eye in the present study. Finally, significant activations in the cerebellum, including key regions for eyeblink conditioning, such as left lobule VI (2 clusters in the left hemisphere: −26, −62, −26; −34, −56, −32) was measured during the 2 trial types, further supporting the position that this structure is important for eyeblink conditioning.

Comparisons between delay and trace conditioning trials showed greater delay-related activity in the bilateral superior temporal gyri (left: −38, −30, 6, 512 voxels; right: 38, −29, 8, 331 voxels) and left claustrum (−34, −14,11, 19 voxels) in all subjects. No significant trace-related activations were observed.

3.2.1.2. Awareness effects.

Table 3 lists significant regions of activation after between-group comparisons of awareness (aware vs. unaware). Subjects classified as aware showed greater activation in the right parietal lobe (48, −46, 35) relative to subjects classified as unaware (Table 3 and Fig. 2A). This awareness effect did not seem to be affected by age as older aware and younger aware subjects also showed greater right parietal activity (older: 48, −44, 33; younger: 47, −46, 35) relative to older unaware and younger unaware subjects. Additional independent analyses included regression models that used individual awareness scores as covariates and showed that greater awareness of the CS-US relationship predicted greater right parietal activity in all subjects (Fig. 2B), further suggesting that neural processing in this region supported greater awareness of the stimulus contingencies. Correlation analyses between individual awareness scores and parietal activity showed a significant positive relationship (r = +0.38, p = 0.007) (Fig. 2C). Fig. 2 further divides aware and unaware participants based on their age (older vs. younger), and one limitation of doing so is that it resulted in a reduction in sample sizes, leading to a decrease in power, and so these data should be interpreted with this consideration.

Table 3.

Significant BOLD differences between aware and unaware subjects

Aware > unaware (all subjects)						Aware < unaware (all subjects)
Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels	Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
L superior temporal gyrus (BA 42)	−64	−29	17	4.1	69	L Precentral Gyrus (BA 6)	−22	−16	51	3.6	22
L superior temporal gyrus (BA 22)	−59	−48	15	3.8	50	L Medial Frontal Gyrus (BA 11)	−1	32	−13	4.2	42
L superior temporal gyrus (BA 13)	−50	−43	19	3.4	13
L Insula (BA 13)	−45	−3	3	3.7	61
L lentiform nucleus/globus pallidus	−25	−15	2	4	36
L lentiform nucleus/putamen	−25	0	6	3.9	84
L superior frontal gyrus (BA 6)	−13	−3	60	3.5	15
R medial frontal gyrus (BA 32)	6	7	47	4.6	114
R medial frontal gyrus (BA 6)	13	3	53	3.6	14
R lentiform nucleus/putamen	25	4	3	4.9	157
R precentral gyrus (BA 6)	41	−10	45	3.5	65
R middle temporal gyrus (BA 22)	49	−45	1	3.5	32
R inferior parietal lobe (BA 40)	48	−46	35	4.3	241
R superior temporal gyrus (BA 42)	54	−33	11	3.3	20
Aware > unaware (older subjects)						Aware < unaware (older subjects)
Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels	Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
L superior temporal gyrus (BA 22)	−59	−44	15	3.7	87	none
L middle temporal gyrus (BA 21)	−57	−60	5	4.2	46
L parahippocampal gyrus (BA 27)	−27	−29	−3	3.6	25	Aware > unaware (younger subjects)
L lingual gyrus	−23	−71	1	4.3	84	Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
L parahippocampal gyrus (BA 34)	−14	−11	−15	3.7	11	R lentiform nucleus/putamen	27	4	0	3.4	15
L medial frontal gyrus (BA 6)	−9	−5	61	3.7	23	R parietal lobe (BA 40)	47	−46	35	3.8	128
L medial frontal gyrus (BA 6)	−9	−1	53	3.5	25
R medial frontal gyrus (BA 32)	4	5	46	5.3	93	Aware < unaware (younger subjects)
R lingual gyrus (BA 18)	3	−76	−5	3.7	25	Brain structure (neocortex, p < 0.001)	x	y	z	SPM{Z}	Voxels
R medial frontal gyrus (BA 6)	17	2	57	4.5	133	L superior temporal gyrus (BA 38)	−30	11	−33	3.9	26
R lingual gyrus (BA 18)	15	−69	−4	3.7	54	L medial frontal gyrus (BA 9)	−9	44	32	4.1	33
R thalamus	21	−12	11	3.7	15
R parahippocampal gyrus (BA 19)	23	−56	−2	3.7	17
R parietal lobe (BA 40)	48	−44	33	3.3	25
R superior temporal gyrus (BA 22)	58	−33	11	4.1	214
R middle temporal gyrus (BA 21)	51	−49	4	3.7	58
R parietal lobe (BA 40)	58	−44	32	3.3	15

Talairach coordinates of activation maxima (Talairach and Tournoux, 1988) as a function of awareness and age.

Key: BOLD, blood oxygenation level dependent.

Fig. 2.

Four independent analyses characterizing the neural correlates underlying CS-US awareness. (A) Magnitude of correlation between learning-related activations and level of awareness for all, younger, and older adults. Higher awareness scores predicted increased activation in the right parietal lobe. (B) A group comparison between aware and unaware participants (based on post-experimental questionnaires) also revealed greater activation in the parietal lobe by aware participants. (C) Scatterplot shows a positive relationship between parietal lobe activity and individual awareness scores. (D) A priori structural ROI analyses using parietal lobe coordinates that most commonly overlapped in lesions in neglect patients (Mort et al., 2003) showed greater activation in this region for all, younger, and older adults. *p < 0.05. Abbreviations: CS, conditioned stimulus; ROI, region of interest; US, unconditioned stimulus.

3.2.1.3. Age effects.

Table 4 lists significant regions of activation after between-group comparisons of age (older vs. younger). Older subjects showed greater activation in multiple neocortical structures, including the right parietal lobe (43, −6, 50) (Fig. 3A), whereas younger subjects failed to show significantly greater activity in any region. Younger and older subjects showed a comparable number of behavioral CRs (t(47) = 0.96, p = 0.34), suggesting that brain activation differences between the 2 groups were not due to a motor performance effect.

Fig. 3.

Effects of age during eyeblink classical conditioning in the parietal and medial temporal lobes. (A) Older adults demonstrated greater activation in the parietal lobe relative to younger adults during learning. (B) A priori structural ROIs based on probabilistic maps of the MTL region (Amunts et al., 2005) were used to show that the medial temporal lobe and subregions were more active in older adults relative to younger adults during conditioning. Abbreviations: CA, cornu ammonis; EC, entorhinal cortex; FD, fascia dentata; MTL, medial temporal lobe; PRh, perirhinal cortex; SUB, subiculum.

3.2.1.4. Interaction between awareness, age, and trial type.

Because older adults demonstrated less awareness than younger adults (t(47) = 2.36, p = 0.02; Fig. 1D), fMRI analyses designed to better understand the interaction between awareness and age were performed. These analyses revealed main effects of age and awareness in the right parietal region. In addition, a triple interaction (aware × age × trial type) was observed in the right parahippocampus (p < 0.001), characterized by high levels of activation in older, aware adults receiving trace conditioning (Fig. 4A). Furthermore, an age × awareness interaction was observed in the left parahippocampus, right hippocampus, and left superior temporal gyrus, revealing high levels of activation in older, aware subjects receiving either delay or trace trial types (p < 0.001; Fig. 4B–D).

Fig. 4.

Neural correlates of CS-US awareness as a function of age. (A) Only older, aware adults receiving trace conditioning showed greater activation in the right parahippocampus (p < 0.001). (BeD) Only older adults who were aware of the CS-US relationship showed greater activity in the left paraphippocampus, right hippocampus, and left superior temporal gyrus. (E) Activation maps showing main effects of awareness (red; p < 0.001) and age (green; p < 0.001). Yellow colors indicate areas of overlap between awareness and age. Abbreviations: CS, conditioned stimulus; US, unconditioned stimulus.

3.2.2. Regions of interest analyses

3.2.2.1. Age and awareness effects.

To further examine the role of the parietal cortex in neglect and contingency awareness, the structural ROI analysis based on coordinates of overlapping lesions in neglect patients (Mort et al., 2003) was used to sample individual subject activations. This analysis further corroborated the parietal/awareness effect by revealing greater fMRI parietal activity in aware subjects relative to unaware subjects in both older and younger subjects (Fig. 2D). Using a priori structural ROIs based on probabilistic maps of the MTL region (Amunts et al., 2005), analyses showed that older subjects showed greater activation in the MTLs and various subregions (Fig. 3B).

3.3. Results from tDCS study

To corroborate the fMRI findings and to further investigate the role of the parietal lobe as it relates to contingency awareness, we applied tDCS to this region during eyeblink conditioning. The timeline of procedures is shown in Fig. 5A. Sixteen naïve subjects (24.3 ± 0.7 years) were randomly assigned to receive cathodal or sham tDCS. Localization was based on fMRI activations (Fig. 5B), and electrodes were applied over these coordinates while subjects received delay eyeblink conditioning trials (see Section 2). After concurrent stimulation and conditioning, subjects’ awareness was assessed using postexperimental questionnaires. Although no significant group differences in CR production were measured (t(14) = 0.13, p = 0.90), the cathodal group reported being less aware of the CS-US relationship in comparison to the sham group (t(14) = 2.57, p = 0.02; Fig. 5C). Critically, there were no significant differences in correctly answered movie question (t(14) = 1.53, p = 0.15), suggesting that the stimulation did not produce a general learning impairment. These findings complement the fMRI results and strongly suggest that the parietal region is highly involved in the conscious processing of CS-US relationships.

Fig. 5.

Effects of cathodal transcranial direct current stimulation (tDCS) over the right parietal lobe on CS-US awareness during delay conditioning. (A) Timeline of events during the tDCS experiment. Participants received a line bisection test before conditioning and received either cathodal or sham stimulation during conditioning. After conditioning and tDCS, a questionnaire probing their awareness of the CS-US relationship and movie content was administered, which was followed by a poststimulation line bisection test. (B) Placement of the center of a 5 × 5 cm tDCS sponge (red circle) was based on fMRI activations (shown in yellow; Fig. 2A) and normalized to each participant’s scalp. (C) No differences in eyeblink CRs were found between the cathodal and sham group (left graph). However, participants receiving cathodal stimulation over the right parietal lobe demonstrated less awareness of the CS-US relationship than those receiving sham stimulation (right graph). Abbreviations: CR, conditioned response; CS, conditioned stimulus; US, unconditioned stimulus.

4. Discussion

This investigation sought to characterize brain activity mediating the interactions of contingency awareness, aging, and delay and trace conditioning using fMRI and tDCS. Major findings include the following: (1) older adults were significantly less aware of the relationship between the CS and US than younger adults, and this result was not attributable to differences in sensitivity to the CS or US, or to general learning deficits, as evidenced by the lack of group differences for correct answers about the movie. (2) Greater right parietal fMRI activation was associated with higher levels of contingency awareness for all subjects receiving either delay or trace conditioning. (3) Cathodal tDCS over the right parietal lobe led to lower levels of awareness in younger subjects without disrupting CRs. (4) Older adults exhibited hyperactivations in the parietal and MTLs, despite showing no conditioning deficits. (5) MTL regions were differentially recruited based on awareness, age, and conditioning trial type.

Prior studies have indicated a link between contingency awareness and trace conditioning and between trace conditioning and MTL structures (Cheng et al., 2008; Clark and Squire, 1998). Because of these associations, and because aging has been shown to impair MTL functions (Daselaar et al., 2006; Dennis et al., 2008), one might expect that age-related decreases in awareness would manifest neuronally as abnormal MTL activations and would preferentially affect only trace conditioning. Our results, however, indicated that age-related decrements in awareness are observed for delay as well as trace conditioning and that right parietal cortex is involved in the brain circuitry for contingency awareness (Fig. 2). Main effects of both awareness and age were observed in the right parietal cortex, but no interaction of either of these variables with conditioning type was seen. These present findings suggest that contingency awareness, as supported by the parietal cortex, may not be uniquely required for trace conditioning, which is a position supported by multiple delay conditioning behavioral investigations using a variety of procedures including differential conditioning (Knuttinen et al., 2001), conditional discrimination (Bellebaum and Daum, 2004), and directed-attention (Weidemann et al., 2016). This also suggests a model in which the parietal cortex serves as a substrate for contingency awareness and that age-related disruption of this region is sufficient to impair awareness in older subjects regardless of conditioning type. Consistent with this model, we found that artificially disrupting the right parietal cortex with cathodal tDCS was able to impair awareness in healthy young subjects performing delay conditioning, a protocol that does not depend on MTL function.

Several eyeblink conditioning studies have investigated the relationships between awareness, the MTL, and trace conditioning (Clark and Squire, 1998; Manns et al., 2001, 2000a,b; Smith et al., 2005). In the first of these studies, Clark and Squire (1998) reported that MTL amnesics failed to acquire trace conditioning. This was attributed to an inability to access awareness after damage to the hippocampus, as healthy controls showed intact trace conditioning only if they were aware. They extended these findings by showing that awareness developed concurrently with trace CRs and was also necessary for both differential and single cue trace conditioning but not delay conditioning (Manns et al., 2000b, 2001; Smith et al., 2005). Our previous study (Cheng et al., 2008) has shown increased activation in the right MTL in trace compared with delay conditioning, and as observed in Fig. 4A, we observed a region in the right MTL that exhibited enhanced activation for trace conditioning in older aware subjects. Importantly, the awareness and trace conditioning relationship proposed by Squire et al were also collected from older participants (mid to late 60s) (Clark and Squire, 1998; Manns et al., 2000a,b). Thus, the development of trace conditioning may also uniquely rely on MTL activation and be accom-panied by contingency awareness, and we hypothesize that this awareness information is supplied by the parietal cortex.

The main debate on the role of awareness and conditioning focused on its necessity during delay conditioning, with both positions agreeing that it is necessary for trace conditioning. Hence, the lack of an awareness effect on behavioral measures of trace conditioning in the present study was surprising. Interestingly, several studies suggest that trace conditioning can occur outside consciousness and awareness. Patients with varying levels of consciousness (vegetative and comatose patients) and healthy sleeping individuals could demonstrate successful trace conditioning (Arzi et al., 2012; Bekinschtein et al., 2009; Juan et al., 2016). Furthermore, awake individuals showed rapid amygdala responding during trace conditioning to unperceived faces, suggesting that trace conditioning is possible without awareness (Balderston et al., 2014).

The locations of the awareness-related right parietal (Fig. 2A) and right parahippocampal (Fig. 4A) activations found in the present study are strikingly similar to the regions that have been found to result in neglect syndromes in middle cerebral artery and posterior cerebral artery stroke patients, respectively (Mort et al., 2003). The interaction of these 2 regions with each other may be particularly important, as Thimm et al. (2008) found that functional activation increases in the right parietal cortex, and decreases in the right parahippocampus were associated with recovery in neglect patients. These patients typically present with a lack of awareness for stimuli presented to the contralesional side of space. A pattern of greater parietal and lower MTL activity was exhibited by our younger subjects (who demonstrated greater awareness), whereas hyperactivation in both regions was measured in our older subjects (who demonstrated lower awareness), suggesting that hyperactivation in these 2 structures may be detrimental to contingency awareness. Furthermore, subjects showing decreased parietal lobe activity demonstrated a lack of contingency awareness. Consistent with this finding, subjects receiving cathodal tDCS over this region were less aware than those receiving sham stimulation, and decrements in awareness were found to correlate (r = –0.54, p < 0.05) with the degree of rightward shift in a line bisection test, a test commonly used to assess neglect. This raises an interesting question as to whether the age-dependent reduction of CS-US awareness is a manifestation of some form of naturally occurring age-related neglect. Consistent with this notion, 2 studies have found that older subjects exhibit abnormal (right shifted) line bisection test results compared with younger subjects (Benwell et al., 2014; Fujii et al., 1995). If reduced CS-US awareness is an indicator of underlying neglect, the parietal and parahippocampal hyperactivations found in our older subjects could provide a biomarker for dysfunction in these regions.

Neuromodulation methods have been used in both healthy and patient populations to study the phenomenon of neglect. In healthy subjects, parietal transcranial magnetic stimulation can induce extinction of contralateral visual stimuli during a simultaneous double stimulus presentation (Pascual-Leone et al., 1994) and produced a significant rightward bias in symmetry judgments of pre-bisected lines as compared with basal and sham repetitive transcranial magnetic stimulation conditions (Fierro et al., 2000). In the tactile modality, right parietal transcranial magnetic stimulation after cutaneous stimulation has produced deficits in detecting either ipsilateral or contralateral stimulation (Oliveri et al., 1999). Two studies have shown that anodal tDCS over the right parietal cortex can reduce visual neglect symptoms in patients with right parietal stroke and contralateral neglect (Ko et al., 2008; Sparing et al., 2009). Sparing et al. (2009) showed that in normal subjects, tDCS could enhance or impair performance depending on stimulation parameters (anodal vs cathodal) and stimulated hemisphere. Finally, simultaneous cathodal and anodal tDCS over the right and left parietal lobes produced stronger and earlier neglect-like effects compared with cathodal stimulation alone (Giglia et al., 2011). These studies indicate that neuromodulation can not only replicate deficits seen in right parietal neglect but can also be used to improve performance. In the present study, cathodal tDCS was seen to produce a decrement in CS-US contingency awareness in young subjects that was correlated with neglect-like effects (line bisection). Future studies can address whether anodal tDCS stimulation is capable of enhancing awareness in older subjects and altering parietal and parahippocampal activation patterns to more closely resemble those seen in younger individuals.

One methodological consideration in the present study is the manner in which awareness was assessed. We used a questionnaire designed to probe subject’s knowledge of the CS-US relationship after the conditioning session. This technique has been used in prior studies (Cheng et al., 2008; Clark and Squire, 1998, 1999; Manns et al., 2000a) but is not without disadvantages, as it may not accurately reflect participants’ awareness of the contingencies during conditioning (due to forgetting). An alternative procedure is to require participants to provide an online US expectancy rating on a trial by trial basis during conditioning. One disadvantage of this technique is that it directs participants’ attention to the US, which may have an unintentional effect on awareness and conditioning levels. Another consideration in the present study is that delay and trace conditioning trials varied in terms of stimulus durations (1350 ms for delay and 250 ms for trace) and collapsing across these 2 trial types may not have accounted for CS saliency effects. However, these trials were matched for interstimulus intervals (1250 ms) and elicited comparable conditioning, suggesting saliency did not differ between the trial types.

In summary, this is the first investigation to use fMRI and tDCS to identify the neural circuitry related to contingency awareness as a function of age. The findings support that the parietal cortex serves as a substrate for contingency awareness and that age-related disruption of this region is sufficient to impair awareness. It is speculated that a naturally occurring mild neglect could manifest as one ages, resulting in lower contingency awareness by our older subjects. Finally, contingency awareness and processing in the parietal region can be modulated by tDCS, which represents a first step in developing a treatment for disorders resulting from parietal damage.