In brief, for past tense production, STN-DBS (with medications) led to lower accuracy at regulars but not irregulars, and to slower response times at regulars than irregulars, as compared to both healthy controls and patients with just medications, who did not differ from each other across assessment sessions. In contrast, for object naming, STN-DBS improved response times (with no changes in accuracy) for the naming of manipulated but not non-manipulated objects, as compared to the other two subject groups, who again did not differ from each other.
These results cannot be explained by various confounding factors, including subject-level variables such as age, education, handedness, or language history, or item-level variables such as frequency or word length, since these and other factors did not differ between groups or conditions or were controlled for statistically. The findings also cannot be explained by speed/accuracy trade-offs: for past-tense production, accuracy and response times showed the same performance patterns, while for object naming, accuracy showed no effects rather than the opposite effect as response times.
The finding that STN-DBS improved naming manipulated (but not non-manipulated) objects in early PD is consistent with previous studies showing that STN-DBS leads to motor improvements, as tested by physical motor outcomes, in advanced PD. The results suggest that previously observed STN-DBS related motor improvements in PD may be due at least partly to improvements in higher-level motor knowledge or processing, and not just to improvements of lower-level motor deficits such as tremor. Moreover, results from this well-controlled task demonstrate that observed motor improvements, or at least aspects related to higher-level motor knowledge and processing, do not seem to be due to confounding factors such as motivation. The findings further strengthen the evidence that naming manipulated objects depends on motor circuits. The results in this study particularly implicate the basal ganglia thalamocortical circuitry that is affected in early PD and the motor portions of this circuitry that are targeted in STN-DBS. It is not yet clear why the improvement in naming manipulated objects was found only with reaction time as the dependent variable, and not with accuracy. However, one possibility is simple ceiling effects in accuracy, since the DBS+ODT patients showed higher accuracy on than off treatment (see ), and on treatment they were quite close to ceiling.
The finding that STN-DBS impairs the production of regular but not irregular past-tenses was not predicted. Rather, we expected that STN-DBS would lead to improvements at regulars (but not irregulars). Crucially, the results differed from our predictions in the direction of the STN-DBS related effect on regulars, but not in which verb type (regulars) and thus which aspect of language (grammar) was affected. This further strengthens links between motor skills and grammatical processing (but not lexical processing), and links to the basal ganglia circuits affected in early PD and targeted by STN-DBS. However, it suggests a more complex relationship between grammar and motor function in the context of STN-DBS than had been anticipated, at least in early PD.
Indeed, the mechanisms underlying the differential impact of STN-DBS on motor and grammar function remain unclear. One possibility is that the pattern may be partly explained by the frequency of the stimulation in the current study (130 Hz), since it has previously been observed that high frequency STN-DBS (as was used here) can lead to motor improvement and cognitive decline, while low frequency STN-DBS enhances cognition but degrades motor function 
. However, the finding that grammar but not lexical processing was negatively affected would still have to be explained. Another possibility is that the decline in performance on regulars is due to negative impacts of STN stimulation on other brain structures involved in grammar. For example, STN-DBS has been found to lead to PET activation changes in various cortical areas as well as in the ventral striatum, concomitant to degraded performance in a cognitive task 
. However, the mechanisms underlying such changes remain unclear. A third possibility is that circuits (basal ganglia thalamocortical loops) that are anatomically close to those targeted by the stimulation suffer from some sort of competition or reorganization, though again the specific mechanisms would need to be determined. Further research on this issue seems warranted.
As discussed above, unlike for the object naming task, the results for the past-tense production task were not predicted. Moreover, they are only partly compatible with previous studies. On the one hand, they are not particularly inconsistent with previous studies of the impact of STN-DBS on lexical processing, since such studies have found that stimulation variously improves, worsens, or does not change performance on these tasks (see Introduction
). On the other hand, the results are different from the two STN-DBS studies of syntax and morphology in advanced PD, which reported improvements on some measures, but no changes on others. One possibility is that some measures show improvements or no change in response to STN-DBS in PD, while others show declines. Another is that different patterns are found in advanced and early PD. A third possibility, however, is that carefully controlling for multiple task, subject- and item-level factors, as was done in the present study, might reveal the actual impact of STN-DBS on grammatical, as well as lexical, processing. Further studies should clarify these issues.
This study has several limitations. First, it is based on a relatively small sample size – though not particularly small compared to other studies of STN-DBS. Second, both the DBS+ODT and ODT patients were always tested on treatment before being tested off treatment. Given the protocol for the prolonged five day washout period, this order could not be easily reversed. Nevertheless, order effects were minimized by avoiding any item repetition between the sessions, in both tasks. Additionally, the direct comparison of on/off differences between subject groups (e.g., DBS+ODT patients to controls) precluded an explanation of more general order effects, unless they were specific to one of the groups. Third, although the DBS+ODT, ODT and control patients were directly or statistically matched on various factors, they might have differed on others (e.g., depression, general cognitive functioning). However, the design of the study, with the within- as well as between-subject and session examination of matched regular/irregular and manipulated/non-manipulated items, minimizes the likelihood of such factors explaining the observed findings. Fourth, and related to the previous point, UPDRS (Unified Parkinson's Disease Rating Scale) scores were not available, since the present study is part of an ongoing clinical trial, and these scores will not be made available until the termination of the trial. However, the strict inclusion and exclusion criteria (), including the fact that all PD patients were at stage II on the Hoehn and Yahr scale when off medication, ensures relative homogeneity within and between the two patient groups (DBS+ODT and ODT). Fifth, with respect to grammar, this study focused only on English regular past-tenses, and thus generalizations to grammar as a whole must be made with caution. Nevertheless, a large literature has linked the processing of English regular past-tenses to other aspects of grammar, including syntax 
, suggesting that such a generalization may not be unwarranted. However, further studies are needed, ideally with the type of control of potentially confounding factors found in the present study. Sixth, it might seem surprising that the production of regular past-tenses was not impaired, as compared to controls, in the off state, given previous evidence that PD patients can show impairments at producing regular past-tenses 
. However, those findings were observed in PD patients at a relatively advanced stage. Thus the lack of a deficit at regular past-tenses observed here is consistent with the more general pattern that cognitive impairments in PD occur at later stages of the disease.
In sum, this study examined the impact of STN-DBS on well-controlled language and motor tasks in patients with early PD, as compared to patients solely on medication, and healthy controls. STN-DBS impaired aspects of grammatical but not lexical processing, while improving aspects of motor function, as compared to both other subject groups, who did not differ from each other. The findings suggest that STN-DBS does indeed impact language, but may be specific in affecting aspects of grammatical but not lexical processing. The differential impact of STN-DBS on motor (improvement) and grammar (degradation) is surprising, and requires further investigation. The finding that aspects of both motor skills and grammar (but not lexicon) were affected by STN-DBS supports links between grammar and motor skills, and procedural memory more generally.
Although the STN-DBS related decline in grammar was found both in accuracy and response times, and thus is not trivial, the size of the observed effect does not seem likely to substantially impact the quality of life of the patients. Thus it still appears that DBS is a viable therapy for PD 
, as it effectively reduces motor symptoms and does not appear to have a major negative effect on language and other aspects of cognition. Nevertheless, the current results suggest the need for further research examining the impact of STN-DBS on language.