Motor precision deficits in clinical high risk for psychosis

Abstract

Motor deficits appear prior to psychosis onset, provide insight into vulnerability as well as mechanisms that give rise to emerging illness, and are predictive of conversion. However, to date, the extant literature has often targeted a complex abnormality (e.g., gesture dysfunction, dyskinesia), or a single fundamental domain (e.g., accuracy) but rarely provided critical information about several of the individual components that make up more complex behaviors (or deficits). This preliminary study applies a novel implicit motor task to assess domains of motor accuracy, speed, recognition, and precision in individuals at clinical high risk for psychosis (CHR-p). Sixty participants (29 CHR-p; 31 healthy volunteers) completed clinical symptom interviews and a novel Serial Interception Sequence Learning (SISL) task that assessed implicit motor sequence accuracy, speed, precision, and explicit sequence recognition. These metrics were examined in multilevel models that enabled the examination of overall effects and changes in motor domains over blocks of trials and by positive/negative symptom severity. Implicit motor sequence accuracy, speed, and explicit sequence recognition were not detected as impacted in CHR-p. When compared to healthy controls, individuals at CHR-p were less precise in motor responses both overall (d = 0.91) and particularly in early blocks which normalized over later blocks. Within the CHR-p group, these effects were related to positive symptom levels (t = − 2.22, p = 0.036), such that individuals with higher symptom levels did not improve in motor precision over time (r’s = 0.01–0.05, p’s > 0.54). CHR-p individuals showed preliminary evidence of motor precision deficits but no other motor domain deficits, particularly in early performance that normalized with practice.

Introduction

Motor deficits are among the first reported signs of psychosis in the premorbid period [1–4]. Later, in the prodrome, these symptoms predict cognitive deficits, functional outcomes, symptom course, and conversion [2, 4–8]. Extant research suggests motor signs may reflect the confluence of early risk for psychosis including genetic [9], early neurological development, familial environment, and/or environmental risk factors [10], making motor deficits a promising indicator of early risk for psychosis. These deficits have also been linked to emerging dysconnectivity in cortical, striatal, hippocampal, and cerebellar neural networks [11–14]. As a result, motor abnormalities may provide an early, sensitive measure of risk for individuals at Clinical High-Risk for Psychosis (CHR-p), aiding in early detection and intervention.

Despite this promise, extant literature primarily focuses on complex behaviors (e.g., gesture performance, hyperkinesia) rather than fundamental components that make up a motor sequence [14–16] or examining such components in isolation [17–19] without consideration for other integral functions in an action sequence. As a result, we have a limited understanding of which principal motor functions are disrupted in this population. Further, motor assessments frequently rely on specialized clinical assessment or specialized motor equipment, limiting the use and translation of motor deficits into a clinical target [20–24]. The current study uses the Serial Interception Sequence Learning (SISL) task as a more comprehensive motor task that provides insight into multiple motor deficits, including sequence-specific accuracy, accuracy adjusted speed, explicit sequence recognition, and motor sequence precision. [25–28]

Although motor deficits are a relatively understudied area of clinical risk for psychosis [11, 14], a growing body of research shows that individuals at CHR-p show deficits in motor sequence accuracy [19], slowing [17], recognition [19], and precision [18, 29], among other abnormalities [13, 30]. These deficits are typically assessed in a single paradigm (e.g., finger tapping [17, 31, 32]) and are typically associated with a single neural domain (e.g., striatal dopamine [31, 33]), which provide limited insight into other motor domains. Additionally, these motor metrics typically examine overall motor performance compared to peers [12, 13, 17, 32] with fewer studies examining differences in motor learning [19, 29]. The few studies on motor learning have demonstrated that CHR-p show reduced late accuracy in pattern recognition [19] and reduced explicit motor precision learning compared to peers [29].

Other perceptual-motor sequence learning tasks, such as the Serial Reaction Time (SRT) task [19], provide insight into motor accuracy, motor pattern recognition, and are associated with hippocampal and dopaminergic mechanisms during early and later improvements to accuracy, respectively [19, 34, 35]. Unfortunately, SRT has no measure of motor precision or motor precision improvements—skills that are associated with cerebellar functions [6, 18, 36]. The cerebellum and cerebellar-related behaviors are often understudied and underappreciated as a region contributing to risk for psychosis [14, 15, 36], but cerebellar deficits appear early [3, 9, 37], which are highly predictive of clinical course [6, 8, 18]. Other tasks have found fine motor precision to be reliably associated with CHR-p status [29], which may reflect the critical contributions of the cerebellum [18, 29, 43]. However, motor precision tasks typically require specialized lab equipment or software [29]. In contrast, SISL provides an accessible, easily administrable, and rapid method of assessing motor precision deficits and precision changes across training via keypress responses to a computer task.

SISL extends SRT by emphasizing precisely timed motor responses rather than immediate stimulus-responses [26–28]. SISL is a relatively more process-pure implicit perceptual-motor sequence learning task, in that full explicit knowledge of the sequence does not influence motor sequence-specific accuracy [28] and is not related to explicit learning or hippocampal activation [26]. Clinical populations with motor deficits and slowing related to reduced striatal dopamine, while being unaffected by cognitive impairments, show impaired implicit motor performance in SISL [25]. Examining these implicit motor behaviors reduces the potential that the measured deficits are due to ineffective strategy [25]. These features are critical to the examination of individuals at CHR-p who often exhibit a defining feature of a drop in global function [44], hippocampal deficits [45–48], and motor slowing [17].

The current study is a preliminary assessment of a novel perceptual-motor sequence learning paradigm to examine several motor domains: motor sequence accuracy, speed, sequence recognition, precision, and precision learning [17–19, 29]. As in prior motor sequence learning work, we expect that this novel task will reveal implicit motor accuracy deficits [19, 49] and reduced motor speed [17, 32] in individuals at CHR-p. We also expect that they will show comparable recognition of the repeated sequences as the sequences are implicit and not expected to be explicitly recognized [19, 28]. Finally, we expect that individuals at CHR-p will show reduced precision in terms of the timing of responses [18, 29].

Materials and methods

Participants

SISL data were acquired on sixty-seven participants (35 CHR-p; 32 Healthy Volunteers (HV)). Participants ranged in age from 18 to 31 (M = 22.1, StD = 2.96; CHR-p: M = 21.93 StD = 2.74 HV: M = 22.26, StD = 3.18). Among these individuals, seven were excluded based on insufficient or incomplete task performance (6 CHR-p, 1 HV). All participants completed a clinical interview, which was used to assess prodromal syndromes for inclusion in the study as an individual at CHR-p. The SCID was administered to all subjects for two reasons: first to rule out formal psychosis and second to assess for other psychiatric disorders with psychotic features. The sample was racially diverse, see Supplemental Table 7. In a set of follow-up analyses, the presence of CHR-p individuals with current medications known to impact cognition and motor performance (antipsychotics-n = 2; stimulants-n = 12) showed no impact on the magnitude and direction of the effects resulting (Supplemental Table 8). No individuals in the healthy control group were on medications. Reported analyses included all available participants.

Clinical assessments

Structured Interview for Psychosis-Risk Syndromes (SIPS) [44] was administered to detect the presence of a prodromal syndrome and to formally assess the severity of attenuated positive symptoms [44]. For inclusion in the study as an individual at CHR-p, individuals had to meet the interview criteria for a prodromal syndrome, and HV had to be free from a prodromal syndrome. Severity was assessed on the symptom frequency, intensity, and content, which was considered in a relevant CHR-p range if the positive symptom dimension received a score of 3 (moderate) to 5 (severe but not psychotic). The total positive and negative symptom score reflects the sum of five positive items. Additional prodromal risk syndrome inclusion criteria included a schizotypal personality disorder diagnosis with cognitive decline or a first-degree relative with a psychotic disorder accompanying a decline in functioning. Exclusion criteria for all participants included diagnoses with a psychotic disorder according to the Structured Clinical Interview (SCID). For healthy control participants, exclusion criteria also included the presence of psychotic disorders in a first-degree relative, which may reflect latent genetic risks. The SCID assessed the presence of a formal psychosis diagnosis for exclusion purposes [50]. All clinical interviews were conducted by graduate trainees in formal clinical assessment training under the supervision of VAM (supervised ratings that are maintained at a reliability of K ≥ 0.80). There were two individuals in the CHR-p risk group and one HV individual that were diagnosed with comorbid depression; follow-up analyses showed that the removal of these three individuals did not change the direction or magnitude of the reported effects.

Serial Interception Sequence Learning (SISL) task

During the SISL task, participants viewed cues moving vertically downward toward one of four labeled targets and were instructed to precisely time a motor response with the key corresponding to the target during the moment of perfect overlap between the moving cue and target (Fig. 1A). Feedback was provided about the accuracy of responses (green flash for correct, red flash for incorrect). A response was scored as correct if the key was pressed within 163 pixels of maximal overlap of cue and target. Cues initially moved down the screen at a rate of 2 s from top to target. An adaptive speed algorithm adjusted the cue velocity individually to target overall task accuracy of 80–85% every 12 trials; increasing by 5% if > 11 responses correct and decreasing if < 9 correct responses; see Supplemental Information (SI). Participants were not instructed that cues followed a 12-item repeating sequence on 80% of trials (4 sequence repetitions per 60 trials with 1 12-item non-repeating segment) during a 1080-trial training phase. The training phase was followed, without warning, by a 1080-trial test phase that contained repetitions of the trained sequence (33%) and repetitions of two novel sequences (33% each). After completing the main training and test phases, participants were given a recognition test to assess explicit knowledge of the repeating sequence acquired concomitantly. Five sequences were shown, including the practiced sequence, and each was rated on a 1–10 scale reflecting confidence that the sequence was seen during training (1 = sure not, 10 = sure seen).

Fig. 1.

A The Serial Interception Sequence Learning (SISL) task. Circular cues (shown in grey) travel vertically down the computer screen toward target zones (open circles) associated with keyboard keys D, F, J, and K. As the cue moves through the bottom target zone, participants attempt make a precisely timed response. Participants are not told that the cues follow a covertly embedded 12-item repeating sequence that is typically learned implicitly. B Motor precision is measured as the absolute linear distance of the cue from the center of the target for correct items within the repeating sequence. Here, cues depict responses that were scored as correct within the allowable scoring window with varying degrees of precision from perfect overlap with their corresponding targets

Sequence-specific accuracy is the difference in accuracy between the trained repeating sequence and novel repeating sequence during the test phase; it increases when accuracy is selectively higher for the practiced sequence.

Overall speed of the task reflects individual differences in task general performance, as overall accuracy is intended to remain constant via the adaptive speed algorithm.

Sequence recognition is the difference in the rating given for the practiced repeating sequence minus the average of the foils, providing a sensitive measure of explicit sequence knowledge during the recognition test.

Motor precision of responses is the absolute linear distance from the cue to the perfect overlap with the target for correct sequence items (Fig. 1B); zero is perfectly precise, and increasing values indicate less precise responses.

Of the initially recruited participants (n = 67), seven individuals were excluded for low performance or incomplete data in the main task (missing > 50% of responses in a block, n = 2; excessive over-responding, n = 2; overall task accuracy < 25%, n = 3). Our final sample included 60 participants (29 CHR-p; 31 HV; 63% female). Additionally, six individuals from the lab setting had incomplete sequence recognition data that was insufficient for analyses, leaving 54 participants for this measure.

Due to the onset of the COVID pandemic during data collection, the task was administered in person to 31 participants (52%) and remotely (online) to 29 participants (48%) using different task implementations. Because we do not have information on the screen size used by participants when completing the task online, all distance measures (precision, speed) are calculated in screen pixels per second. An error in the speed adjustment algorithm in the experiment administration code for the remote participants led to a difference in performance on some task metrics (speed over-accelerated, leading to overall lower task accuracy but similar sequence-specific learning). Follow-up analyses examining the impact of lab settings indicated that there was a substantial difference observed in speed (and overall accuracy) between participants who completed the task in person versus using the online version due to an error in the online version that was associated with overall higher speed and lower accuracy for both CHR-p and HV participants (see SM). No differences were found between groups of participants in either online or in-person groups.

Analytical strategy

Demographics were compared across clinical risk groups (CHR-p, HV); any group differences in demographics were included in the models that compare performance across groups. The SISL task motor metrics (sequence-specific accuracy, overall speed, motor precision) were compared across groups in separate multilevel models, which included fixed effects examining the main and interactive effects of training blocks and diagnostic group while accounting for the fixed effects of setting (lab, online) and the random effects of trial-level data that were clustered within each subject and training block, see Supplemental Table 3 for full model statistics and group means. Given that motor recognition involved relatively few items, group differences in recognition were compared in a simple t-test. The inclusion of the administration context was done out of concern for the difference in overall performance resulting from the speed adjustment error and the fact that the average SIPS scores were slightly different among these two subgroups (See SI for additional analyses and full description of follow-up analyses). For transparency, all analytic code (R.v.3.6.3) [51] was provided in supplemental materials and was independently reviewed by authors (KSFD, ZH, YCH) for accuracy.

Results

Participants

Our final sample included 60 participants (29 CHR-p; 31 HV; 63% female). Additionally, 6 individuals from the lab setting had incomplete sequence recognition test data that was insufficient for analyses, leaving 54 participants for this measure. There was not a significant difference in the ratio of sex by group, χ² = 1.00, p = 0.32. There was also not a significant difference in age between groups, t(65) = 0.43, p = 0.67. There was also not a significant difference in race between groups, χ² = 7.39, p = 0.19. There was no significant difference in years of parental education, t(65) = 0.31, p = 0.76. All participants exhibited robust sequence-specific accuracy (M = 5.2%, SE = 1.0%), indicating sufficient task engagement.

Sequence-specific accuracy

There was no difference in sequence-specific accuracy by group, t(56) = 0.57, p = 0.57, nor any group by block interaction, t(56) = 0.93, p = 0.36.

Overall speed

There was also no group difference in overall speed (overall task difficulty) needed to maintain the target performance level, t(56) = − 0.16, p = 0.87, nor any group by block interaction, t(56) = 0.35, p = 0.73.

Sequence recognition

There was no group difference in sequence recognition of the embedded sequence after training, t(53) = 0.27, p = 0.78.

Motor precision

In motor precision, there was a significant main effect of group (t(51) = −3.05, p = 0.004, d = 0.35 [− 0.162–0.859 95% C.I.]), indicating that the CHR-p group was overall less precise compared to the HV group (Table 1, Fig. 1B). There was also a significant group by block interaction (t(56) = 3.34, p = 0.0015, Fig. 2). In follow-up analyses to describe the group by block interaction, the average response precision at the beginning (first 240 trials) and end (last 240 trials) was compared between groups with a mixed-model 2 × 2 ANOVA. The significant group by block interaction was driven by early learning deficits in the CHR-p group (d = 0.79 [0.261–1.31 95% C.I.]). Precision improved across the practice blocks in the CHR-p group such that their performance was similar to HV in later blocks, F(56) = 9.15, p = 0.004, d = 0.18 (− 0.32–0.687 95% C.I.), Table 1, Fig. 1C. The main effect of group remained significant, indicating that the CHR-p group was overall less precise compared to the HV group, F(56) = 8.83, p = 0.004, Fig. 2.

Table 1.

Motor Precision (in pixels) Group Averages and Effect Sizes

Training Block	CHR-p		HV		Group difference
	Mean	StD	Mean	StD	Cohen’s D
All	39.3	16.96	34.1	12.94	0.35
Early (First 240)	48.08	20.85	34.1	14.38	0.79
Late (Last 240)	34.62	13.55	32.3	12.28	0.18

CHR-p Clinical High-Risk for Psychosis, HV Healthy Volunteer

Fig. 2.

A Average motor precision in CHR-p (red) individuals compared with HV across 60-trial blocks during SISL training. Higher scores reflect less precision as more distance between the cue location and ideal performance when the response was made by the participant. B Throughout early training performance (average of first 240 trials), the CHR-p group exhibited less precise responding (~ 48 pixels from the D target), although with practice, their responses improve in precision to become similar to HV participants in late training (average of last 240 trials, ~ 34 pixels from the J target). CHR-p – Clinical High-Risk for Psychosis; HV – Healthy Volunteers

Motor precision and positive symptoms

For this analysis, the change in precision (average early precision minus late average precision) was the dependent variable in a linear model. Predictor variables included total positive symptoms and a factor reflecting whether the task was administered online or in person. There was a significant effect of positive SIPS score, t(27) = −2.22, p = 0.036, reflecting the fact that positive symptom severity was associated with lower levels of learning across practice, Fig. 3. Individuals with greater symptom severity showed less precision learning over blocks (r’s = 0.01–0.05, p’s > 0.54), while those with moderate and low symptom levels improved in motor precision over blocks (r’s = − 0.29 to − 0.59, p’s < 0.002). There was no reliable effect of the lab setting, p = 0.89, or interaction between lab setting and positive symptoms, p = 0.84, indicating positive symptom severity alone led to differences in motor precision learning amount.

Fig. 3.

Motor Precision Learning by SIPS Total Positive Symptoms Severity. Motor precision learning is the improvement average response precision from early in training (first 240 trials) to the end of training (final 240 trials). The amount of improvement is lower for individuals with higher total positive SIPS symptom scores. This relationship was similar for the subgroup who completed the task online and the group that completed the task in person (online/lab participants graph, see Supplemental Fig. 2)

Discussion

Individuals at CHR-p showed poorer early motor precision on the SISL task compared to peers, exhibiting less precise responses at the beginning of training but improving over the training phase to reach a normalized performance level. However, this learning effect varied across CHR-p participants, such that greater symptom severity was associated with less improvement over training. These findings are consistent with prior research that suggests that cerebellar-dependent motor functions, such as motor precision [29], coordination [2, 4, 9, 37, 52], and postural sway [5, 6, 53], are impacted early in psychosis disease course and may be compensated for with cognitive resources. These findings are consistent with a previous finding that individuals with the greatest symptom severity have greater cognitive, motor coordination, and sequencing deficits [8]. Motor precision and deficits in motor precision performance over time may be useful tools to assess CHR-p risk and severity [39].

There were several motor domains for which a deficit was expected in the CHR-p group [17, 19] but was not detected although prior studies have found that overall motor sequence learning, accuracy, and speed may not be impacted in those at CHR-p [19, 54] or with a psychosis diagnosis [55, 56]. This discrepancy may also be due to task features [27, 28]. In the current study, no mention is made of potential nested patterns within the responses until the end of the task. In previous studies, the subjects were prompted to consider whether there were hidden patterns after every 50-trial block [19, 55], which may have impacted the types of strategies employed by CHR-p individuals [57, 58]. Additionally, previous studies have found that differences in accuracy occur in the rate of improvement over time, but not the overall accuracy or speed [19, 55, 58, 59]. In these cases, motor precision was not assessed separately from accuracy or speed [27]. As a result, the prior findings may have been in part due to the precision errors uniquely detected in the SISL task [28, 60]. Finally, the SISL task captures differences that are driven by early blocks (4 blocks of 60 trials = 240 total trials/1080 trials) [28], while previous tasks occur within a much narrower window of trials (6 blocks of 50 trials total) [19, 58]. This limited window into motor learning may impact estimates of motor learning in the CHR-p group, and other tasks may benefit from an expanded learning window. Taken together, the current study emphasizes the importance of examining precision, an orthogonal measure of motor performance separable from accuracy and speed [27].

The CHR-p group was less precise in motor responses than healthy peers, which was driven by early blocks. Although this early/late motor precision performance distinction is distinct from early (hippocampal associated) and late (dopaminergic associated) motor accuracy effects previously described in the SRT task in mild cognitive impairment [35], Parkinson’s disease [34], and CHR-p [19] groups, similar motor principles may apply. Early motor precision may be associated with early cortico-thalamic-cerebellar circuitry deficits, while slower cortico-striatal/cortico-cortical motor circuitry function may account for later improvements in precision [14, 61–63]. Although this is consistent with human [61, 63] and animal [62] models of early skill motor learning, future studies are needed to directly relate this deficit in individuals at CHR-p to specific neural mechanisms.

Deficits in motor precision and motor precision performance over time were also related to symptom severity. Individuals with high positive symptom levels showed no significant improvement in motor precision over training in contrast to individuals with lower positive symptom levels. This finding is consistent with motor clustering studies that have found that heterogeneity in motor response reflects symptom heterogeneity [8]. Individuals with the greatest symptom severity may have the greatest severity of deficits in neural systems or be less able to compensate for or improve from initial motor precision deficits [38, 39, 49].

The current study has many strengths, but it is also important to note some weaknesses. The current sample size is consistent with extant literature examining motor deficits in CHR-p [5, 19, 30, 53, 64] (n’s = 22–46) and examinations of group differences in SISL performance [25, 26, 60, 65] (n’s = 11–26). However, a larger sample would enable the examination of heterogeneity within the CHR-p sample. It is also noteworthy that the task was completed both online and in person due to concerns about safety during the global pandemic. Indeed, for some aspects of the task, including sequence-specific accuracy and motor precision, the task setting did impact performance similarly across CHR-p and HV individuals due to a difference in task programming but did not impact the differences between groups, the main question of interest. Future studies should examine several variables that were not included in this initial study, including fitness level, body mass index, and working memory assessment, among others. Finally, the current study reflects an initial examination of motor precision and should be considered the first step in a larger body of work to examine how this motor precision deficit might relate to other motor timing [18, 66], variability [29], and precision tasks [39], as well as relevant cerebellar-thalamic-cortical connectivity [6, 18]. As a result, future studies should validate this measure against current gold standard assessments of motor deficits such as neurological soft signs. These future directions may provide targets for early intervention to address deficits in cerebellar networks underlying motor precision deficits [39, 40].

Conclusion

In this initial validation of a novel task, CHR-p individuals showed deficits in motor precision but no other motor domain deficits, particularly in early performance that normalized with practice. This may reflect distinct precision motor learning system engagement over time, specifically early blocks tend to be associated with cerebellar circuits known to be dysfunctional in CHR-p populations [18, 36, 38] (e.g., cognitive dysmetria) [43] in early blocks that may be addressed in later precision blocks that have been associated with cortico-cortical/cortico-striatal systems [61–63]. Within the CHR-p sample, this early motor precision deficit improved over blocks in individuals with lower positive symptoms, respectively, whereas individuals with higher levels of positive symptoms showed less improvement in precision over blocks. This within-group heterogeneity may reflect heterogeneity in symptom severity, including comorbid diagnoses and symptoms like depression [9, 67], but future work should examine the contributions of associated neural network deficits and distinct clinical subtypes of deficits that span motor, symptom, and cognitive domains [8]. Additionally, it should be noted that the current study is a preliminary study of a novel motor task. The conclusions of this study should be made cautiously and be validated in the future with larger, independent samples using Bayesian approaches. Motor precision is a relatively understudied motor domain, yet precision shows great potential to be an early indicator of attenuated psychosis symptoms as well as symptom severity, and the SISL task should be considered for inclusion among standard batteries of motor tasks in psychosis.

Data availability

Data will be made available upon request to VAM only for those analytic purposes covered in the original consent approved by the Northwestern IRB. Analytic code is provided among the Supplemental Materials. SISL Task access and inquiries should be directed to PJR.