DISCUSSION

The present study provides evidence that basal ganglia dysfunction from vascular or degenerative damage can lead to a common pattern of impaired controlled semantic priming not present in matched individuals with cortical lesions or controls. Specifically, NS and PD subjects were able to process lexical ambiguities presented in isolation via automatic and attention-based procedures; however, attentional processing was typically limited to the facilitation of both dominant and subordinate meanings, implying a deficit in selectively engaging the semantic network on the basis of meaning frequency. In the following discussion, the similar results of the control and CL groups are interpreted in terms of models of lexical ambiguity processing, then the NS and PD group findings are interpreted first in terms of a breakdown in attentional engagement and semantic selection and then with reference to possible underlying frontal–subcortical neurocognitive mechanisms.

The control subjects and CL subjects evidenced a pattern of priming consistent with multiple lexical activation at 200 ms ISI, followed by selective facilitation of the dominant meaning at 1250 ms ISI. These results are in keeping with the findings of Simpson and Burgess (1985), who found that the dominant meaning of an ambiguity was selectively facilitated at a very brief SOA (16 ms); however, by 300 ms the subordinate meaning was also significantly facilitated. This pattern of priming suggests that multiple meanings are activated for lexical ambiguities presented in isolation, but the speed of activation for each particular meaning varies as a function of its relative frequency. In view of these observations, the current finding of nonselective meaning facilitation at 200 ms ISI may be considered to reflect the intermediate stage of lexical ambiguity processing, where the less frequent meaning has been allowed sufficient time to be activated to a level above the priming threshold. The intact priming observed for the CL group is also similar to previous reports of normal word-pair priming of ambiguities in individuals with aphasia (Katz, 1988).

Following the multiple meaning facilitation witnessed at 200 ms ISI, subordinate meanings were no longer facilitated at 1250 ms ISI in control subjects and CL subjects. Similarly, Simpson and Burgess (1985) reported that dominant meaning facilitation was maintained beyond 300 ms SOA, while the level of activation for the subordinate meaning was diminished over time until it did not differ significantly from unrelated words at 750 ms SOA. A subsequent experiment by Simpson and Burgess (1985) utilized a neutral prime condition to demonstrate that this reduction in subordinate meaning activation was due to inhibition of the less frequent meaning. It was suggested that following an automatic activation of all related meanings, subjects used the ambiguous prime to actively direct attention toward the dominant meaning, at the expense of subordinate meanings. As the present experiment did not include a neutral condition, we can only speculate that subordinate meanings in the present study were also inhibited through limited-capacity attentional mechanisms rather than simply decaying, based on the consistency of the present findings in the CL and control groups with the findings of Simpson and Burgess (1985) for related and unrelated targets.

The NS and PD subjects showed a pattern of priming that diverged from the normal pattern of multiple lexical activation followed by frequency-based meaning selection. At 200 ms ISI, the NS and PD subjects showed priming for both dominant and subordinate meanings, consistent with previous findings in young normal controls (Simpson & Burgess, 1985), and the present group of controls and CL subjects. This finding suggests an intact automatic activation of all related meanings of an ambiguity upon its presentation. At 1250 ms ISI, priming for dominant and subordinate meanings was maintained in NS and PD subjects, unlike normal controls and CL subjects. At this stage, Simpson and Burgess (1985) suggested that the subordinate meaning is usually inhibited as limited-capacity attentional resources are directed toward the dominant meaning of the ambiguity. As the lack of a neutral condition precluded the true dissociation of facilitation and inhibition, it can only be speculated that NS and PD subjects were unable to focus or constrain attention toward the dominant meaning and actively suppress the subordinate meaning. It should be noted that the pattern of nonselective priming exhibited by the NS and PD subjects at 1250 ms ISI does not represent an absolute failure in attention-based lexical processing per se, as the subjects were able to maintain lexical activation over time, presumably through attentional/strategic processing.

Instead, the aberrant maintenance of subordinate meaning activation suggests an inability to constrain attention within the lexical–semantic network on the basis of the relative meaning frequency, possibly through deficient inhibitory mechanisms. The possible locus of this inhibitory disturbance is illustrated in Figure 1, which shows one proposal for the lexical and semantic structure of ambiguous words, and the facilitatory and inhibitory operations involved in their computation. Briefly, it has been argued that lexical ambiguities are represented by a single lexical node at the entry level, which activates separate meaning nodes at the semantic network or word sense level, which are connected by inhibitory pathways (Balota et al., 1999; Cottrell, 1989; Tanenhaus et al., 1987; but see Balota & Paul, 1996). The presentation of an ambiguity causes an obligatory activation of all word senses via bottom-up excitatory connections, while lateral inhibitory connections between word senses at the semantic level provide the mechanism for selective meaning facilitation, usually on the basis of contextual feedback, or in this case, on the basis of the strength of associations. These inhibitory mechanisms may be passive in the sense that once activation accrues for one meaning representation, it automatically decreases for connected representations, without explicit attentional control (Balota & Paul, 1996). Some form of disturbance in these inhibitory mechanisms within the semantic system may be indicated in the NS and PD subjects.

Potential lexical and semantic structure of lexical ambiguities. Facilitatory and inhibitory mechanisms for ambiguities presented in isolation. Arrowed pathways (i.e. →) indicate facilitatory connections. Segmented pathways (i.e., [bull ]—[bull ]) reflect inhibitory connections. Adapted from Balota et al. (1999).

The implication of the basal ganglia in semantic inhibition appears analogous to a proposed basal ganglia role in motor functions; namely, that the basal ganglia act focally to select appropriate motor programs and inhibit competing motor mechanisms (Mink, 1996). A disturbance in competitive and inhibitory semantic processing has been suggested previously in subjects with PD on the basis of word production performance (Gurd & Oliveira, 1996). Specifically, Gurd and Oliveira (1996) found that individuals with PD evidenced difficulties in selecting an appropriate word from semantic competitors in a word search task, suggesting difficulties in semantic inhibition. Interestingly, a neural network simulation of semantic processing in PD also demonstrates semantic processing deficits in terms of aberrant semantic inhibition between competing meanings of lexical ambiguities (Watters & Patel, 2002); however, PD-like errors occurred due to increased inhibition for ambiguities with similarly frequent meanings but did not effect unequibiased ambiguities; a finding which is clearly at odds with the present data. The apparent inability of NS and PD subjects to focus attention within the semantic network is suggested by previous evidence from a lexical decision priming paradigm where PD patients experienced difficulties in semantic set shifting (McDonald et al., 1996). This finding is also consistent with difficulties experienced by PD subjects in focusing on salient information and ignoring irrelevant information (Brown et al., 1997; Levin et al., 1989). The present data provide further evidence that individuals with basal ganglia dysfunction have difficulties in selecting meanings, whether on the basis of meaning frequency, as shown presently, or through the attention-based integration of lexical, sentential or discourse level contextual constraints (Copland et al., 2000a, 2000b, 2001).

How may the present data be accommodated by theories and conceptions of subcortical language function? At a general level, the current findings are in keeping with the view that the basal ganglia are involved in lexical–semantic operations (Cappa & Vallar, 1992; Wallesch & Papagno, 1988). The evidence of intact lexical–semantic representations following basal ganglia dysfunction is in agreement with the common assumption that such information is located or stored cortically rather than in subcortical locations (Crosson et al., 1997). The performance of the NS and PD cohorts implies that the basal ganglia are not significantly involved in the automatic accessing of lexical–semantic information, but may play a supportive role in the attention-based processing of this information. This proposal is further endorsed by recent neuroimaging data in healthy individuals, indicating increased striatal activity during controlled but not automatic semantic priming (Rossell et al., 2001) and findings of increased basal ganglia activity during various semantic judgment tasks which are assumed to involve or allow for controlled processing (e.g., Abdullaev et al., 1998; Binder et al., 1997; Mummery et al., 1998; Price et al., 1997).

The present findings are not consistent with the argument that language processing is not affected in PD individuals due to functional reorganization of the language systems (Wallesch & Papagno, 1988). Instead, the current data add to mounting evidence that various aspects of language processing are indeed compromised in PD, including semantic priming (e.g., Arnott et al., 2001), lexical ambiguity resolution in sentences (Copland et al., 2000a, 2001) and sentence processing (Grossman, 1999). The same pattern of priming for NS and PD groups is also at odds with the view that language disturbances following NS lesions are due to white matter pathway disruption causing cortico–cortical or thalamocortical disconnection (Alexander et al., 1987). Instead, the inability to modulate meaning activation for lexical ambiguities in PD and NS individuals is consistent with the general position that frontal–striatal systems may contribute to lexical–semantic processing, but the present findings do not support the view that this role is restricted to the release or selection of lexical items for production as postulated by previous models of subcortical language function (Crosson, 1985; Wallesch & Papagno, 1988).

Interestingly, a more recently proposed model of thalamic language function (Crosson, 1999; Nadeau & Crosson, 1997) may be of relevance to the current findings. According to this model, the thalamus acts via the frontal lobes to selectively engage aspects of the lexical–semantic network, which serves to heighten the difference in activation between representations during cognitive operations. This mechanism has been developed primarily with lexical output functions in mind, where an appropriate response is selected from competing representations for naming purposes. It may be argued that the selective activation of a dominant meaning via attentional semantic processing mechanisms represents an analogous competitive situation within the semantic network. Given the postulated structure of lexical ambiguities with regards to semantic–level inhibitory mechanisms (see above), the present finding of impaired meaning selection via attentional operations in the subcortical groups may be viewed as a disruption of “semantic engagement” mechanisms due to nonthalamic subcortical disruption. This interpretation suggests the need to further consider the neural underpinnings and functional scope of the selective engagement model of thalamic language function.

Although the neural basis of the present deficits remains a point of contention, it is argued that a breakdown in proposed frontal–subcortical attentional and strategic operations via various pathophysiological mechanisms provides a parsimonious explanation for the dissociation between spared automatic lexical processing and compromised attentional/strategic processing in the NS participants. This position is strengthened by the observation that controlled semantic processing served to discriminate between the NS and CL groups, and was disturbed in a similar manner in the NS and PD groups.

In addition to previously cited neuroimaging evidence of possible basal ganglia involvement in controlled semantic processing (e.g., Rossell et al., 2001), the present findings are consistent with a disruption of frontal–subcortical circuits. A disturbance of the anterior cingulate loop functions may account in some way for the breakdown in attentional control of meaning activation in the PD and NS individuals. Posner and DiGirolamo (1998) proposed that the anterior cingulate circuit including the basal ganglia is involved in mechanisms of executive attention which are invoked under certain conditions where automatic or routine processes are inadequate, including the attentional control of semantic activation through various “top-down” processes. One such situation is the presentation of a lexical ambiguity in context, where the anterior cingulate is suggested to be involved in the development of expectancy-based processes responsible for bringing a single contextually appropriate meaning to consciousness (Posner & DiGirolamo, 1998). Early et al. (1989) also proposed the involvement of the anterior cingulate in the attentional control of the semantic relations of words, citing the resolution of lexical ambiguities as a salient example of a situation where anterior cingulate attentional mechanisms may be invoked to select one meaning from a range of competing candidates. Based on the present findings, it is speculated that this postulated function may also extend to the selection of meanings on the basis of meaning frequency in the absence of disambiguating context. There is already evidence of increased activity in the anterior cingulate associated with other relevant tasks involving attention-dependent semantic processing (Frith et al., 1991), and controlled semantic priming (Mummery et al., 1999; Rossell et al., 2001).

The current findings are also consistent with Gold et al.'s (1997) proposal that a disruption in the dorsolateral prefrontal loop via striatocapsular lesions may prevent the formation of lexical or semantic strategies and as a consequence disrupt activation of the semantic network. According to the semantic activation hypothesis (Gold et al., 1997; Gonzalez Rothi, 1990), semantic or lexical strategies are developed to selectively engage and activate aspects of the semantic network. Gold et al. (1997) reported impaired semantic strategy formation in a patient with bilateral striatocapsular lesions and proposed that a disruption in the DLPFC loop may prevent the formation of semantic strategies and as a consequence disrupt the selective engagement of the semantic network. The inability of the PD and NS subjects to selectively facilitate dominant meanings at the long ISI may also be conceptualized as a failure to sustain a semantic strategy that guides meaning selection within the semantic network on the basis of meaning frequency, although it should be noted that the selection of dominant meanings within this paradigm appears quite obligatory (Simpson & Burgess, 1985).

Finally, Kischka et al. (1996) proposed that dopaminergic systems, which include frontal–striatal projections, may serve to modulate activation levels within semantic networks by amplifying stronger signals and dampening weaker signals (Kischka et al., 1996). This argument was supported by the observation that the ingestion of L-dopa by healthy subjects modulated semantic priming effects by decreasing priming of indirect associates. Based on the present data, it may be speculated that the proposed dopaminergic regulation of signal-to-noise ratio within the semantic network may be expressed through frontal–striatal systems, and that the notion of attenuating stronger and weaker signals is compatible with the selection of dominant meanings and the inhibition of subordinate representations within semantic memory. While this proposal is plausible, it is at variance with the argument advanced by Kischka et al. (1996) that the dopaminergic modulation of semantic network activation most likely occurs through mesocortical projections.

In summary, the present study provided a novel comparison of individuals with cortical lesions and individuals with vascular or degenerative basal ganglia damage on a semantic priming task. The significant difference in the age of the PD group and the control group is a limitation of the present study, however, this does not impact upon the major findings presented. The similar performance of the NS lesion and PD groups on the lexical ambiguity priming task suggest that basal ganglia dysfunction interrupts the attention-based selective engagement of the semantic network on the basis of meaning frequency. Although this finding is at odds with traditional theories of subcortical language function, it is in keeping with recent neuroimaging data and current conceptions of frontal–subcortical mechanisms supporting the controlled processing of semantic information.