Semantic interference during blocked-cyclic naming: Evidence from aphasia

Abstract

Nonaphasic speakers are known to take longer to name pictures when they are blocked by semantic category and repeated multiple times. We replicated this “semantic blocking effect” in older controls and showed that in aphasia, the effect is manifested in increased error rates when naming semantically homogeneous, compared to mixed blocks. We further showed that semantic blocking affects Broca’s aphasics more than a matched group of NonBrocas, and that the effect increases with repetition of the blocked sets. Error analysis undermines the inhibition-based account of the blocking effect by showing that errors arise from competition among increasingly activated items within the homogeneous set. The consequent slowing of naming latencies is due at least in part to the intervention of a controlled selection mechanism, and the disruption of this mechanism in anterior aphasia accounts for the increase in error vulnerability.

Introduction

The task of naming pictures invokes semantically driven lexical retrieval, which, by most accounts, is a competitive process (Dell, 1986, Humphreys et al., 1988, Levelt, 1989, Levelt et al., 1999, Roelofs, 1992, Stemberger, 1985, Wheeldon and Monsell, 1994). The occurrence of semantic errors in standard picture naming is prima facie evidence that retrieval is semantically based (e.g. Dell, Schwartz, Martin, Saffran, & Gagnon, 1997), while experimentally elicited interference effects attest to co-activation of a semantic cohort whose members compete for selection (e.g. Vitkovitch & Humphreys, 1991). Among the paradigms that elicit semantic interference in naming are the picture–word interference task (Glaser and Dungelhoff, 1984, Glaser and Glaser, 1989, Lupker, 1979), primed, speeded picture naming (Vitkovitch and Humphreys, 1991, Vitkovitch et al., 1993, Experiment 2), alternating picture naming and naming to description (Moss et al., in press, Wheeldon and Monsell, 1994) and semantic blocked naming. This paper is concerned with the nature of the interference effect in semantic blocked naming (hereafter, blocked naming).

The blocked naming paradigm manipulates the context in which pictures to be named appear, with successive trials depicting items from the same or related semantic categories (homogeneous context: e.g., TRUCK, CAR, BIKE),¹ versus mixed-category items (mixed context: e.g., TRUCK, FOOT, DOG). The basic effect is that targets are named more slowly in the homogeneous context, when pictures are named once (Brown, 1981, Kroll and Stewart, 1994), or multiple times pseudorandomly (Damian, 2003, experiment 2; Damian et al., 2001, Maess et al., 2002). In a variant of this paradigm, which we refer to as blocked-cyclic naming, sets of items (homogeneous or mixed) are named in succession multiple times, e.g., the animal set is presented once, then repeated again in a different order for some number of cycles (Belke et al., 2005, Hodgson et al., 2003, Pickard et al., 2003). This blocked-cyclic manipulation has the potential to elucidate how interference emerges across repetitions (e.g. Belke et al., 2005). We used it here to explore the build-up of blocking interference in persons with aphasia.

Because blocked naming affords an opportunity to manipulate interference during word retrieval, variants of the paradigm have been used to investigate causes of word retrieval deficits in aphasia. Two patients have been reported who, when administered a version of semantic blocked-cyclic naming, produced significantly more errors on homogeneous compared to mixed blocks.

The first of these patients (FAS, reported in McCarthy & Kartsounis, 2000) exhibited nonfluent propositional speech with agrammatism, secondary to tumor. He also exhibited an unusually variable form of anomia, which proved to be sensitive to blocking context, cyclic repetition, and presentation rate. His naming accuracy was lowest for homogeneous blocks presented with a short response–stimulus interval (Rate 1 s vs. 10 s), and accuracy in this condition declined across cycles.

The second case, BM, was one of two reported by Wilshire and McCarthy (2002). BM exhibited nonfluent propositional speech without agrammatism, secondary to left cerebral vascular accident. BM was run on the blocked-cyclic naming task with 2 or 3 s to name each picture. At the faster rate, he made about 45% errors in the homogeneous condition, compared with 15% errors in mixed. At the slower rate, the difference between context conditions was not statistically reliable. Wilshire and McCarthy’s second subject had anomia secondary to a posterior (left temporal) lesion. His naming was unaffected by the blocking manipulation or by the interaction of blocking with presentation rate or repetition cycle.

While these two studies suggest selective vulnerability to the blocking effect in patients with anterior forms of aphasia, the handful of other patient studies that utilized blocked naming have mostly produced negative effects for aphasics of both anterior and posterior types (Gotts et al., 2002, Hodgson et al., 2003; negative in one of two patients tested, Lambon Ralph, Sage, & Roberts, 2000; Schwartz and Hodgson, 2002, Williams and Wright, 1985). Thus, the literature raises questions about the generality of the findings from patients FAS and BM and the significance of the fact that their aphasias were of the anterior type (i.e., featuring nonfluent propositional speech with or without agrammatism). Our study explores these issues using a contrastive group study design.

It is clear that the blocking effect hinges on semantic relatedness. First, the size of the effect tracks the degree of relatedness of items in the homogeneous blocks (Vigliocco, Vinson, Damian, & Levelt, 2002). Second, the effect can generalize to semantically related items that were not part of the response set (Belke et al., 2005). Third, the effect persists even when the visual similarity of homogeneous- and mixed-block pictures is equated, which argues against a visual-conceptual locus (Belke et al., 2005, Damian et al., 2001, Vigliocco et al., 2004). Finally, the effect is present with picture naming, which involves semantically mediated retrieval, but absent with word naming, which can bypass semantics (Damian et al., 2001, Kroll and Stewart, 1994).

Additional evidence from Damian et al. (2001) attests to the lexical basis for the semantic blocking effect. In an experiment with German speakers, these investigators showed that whereas there was no semantic blocking effect when picture names were simply read aloud, the blocking effect emerged when participants instead produced the name together with its gender-marked determiner. On the notion that the choice of determiner in German is governed by the noun’s lexical entry, this finding constitutes strong evidence that the blocking effect is lexically based and, together with the aforementioned evidence for semantic involvement, justifies locating the effect at the lexical-semantic processing level (hereafter, lexical level) (Belke et al., 2005, Damian, 2003, Damian et al., 2001, Maess et al., 2002).

Broadly speaking, a naming target can suffer blocking interference from the heightened excitation of semantic competitors (“too much excitation”) or from the target’s reduced availability due to inhibitory processes of one form or another (“too much inhibition”).

Inhibitory processes in the form of post-selection inhibition were invoked to explain patient FAS’s performance on the blocked-cyclic naming paradigm (McCarthy & Kartsounis, 2000). McCarthy and Kartsounis claimed that this inhibitory process, which purportedly serves sequential selection by imposing a brief refractory state on the selected representation (e.g. Mackay, 1982), had become pathologically protracted in FAS at the level of lexical nodes. The blocking and cycling manipulation exaggerated this condition, they claimed, by causing the inhibition to build across cycles and to spread among related lexical nodes, making it difficult for FAS to recall any appropriate word and causing him to become “jammed and mute” (p. 495).

Belke et al. (2005) proposed a different, purely excitatory, account of the blocking effect, based on the Weaver++ model of lexical access (Levelt et al., 1999, Roelofs, 1992, Roelofs, 1997). They proposed that the blocked-cyclic paradigm imposes a refractory state on the current target by exaggerating the competition between the target and its category-coordinates: “When repeatedly naming objects from homogeneous object sets, (residual) activation accumulates in the semantic system within a small set of related lexical concepts and their shared category and feature nodes.” (p. 687). During lexical retrieval, this high activation density gets mapped to the lexical level, creating a high competition situation. For example, during lexical retrieval of successive animal names, activation accrues in the semantic node representing the category ANIMAL and in the lexical concepts representing animal exemplars. When this residual activation passes to the lexical level, competition is heightened among the corresponding lexical nodes. This lengthens the time needed for a target node to exceed the critical difference threshold, which considers the target’s activation level relative to the sum of the activation of all competing lexical nodes (see also Wheeldon & Monsell, 1994).

Whereas “too much inhibition” and “too much excitation” accounts address the mechanism of semantic interference, what we will call the “executive selection account” speaks to lexical selection in the face of semantic interference. An example is Wilshire and McCarthy’s (2002) explanation for BM’s deficit in blocked naming (see also Robinson, Blair, & Cipolotti, 1998). They propose that BM’s anterior lesion compromised the workings of a selection mechanism that is extrinsic to the mental lexicon and that operates in a controlled fashion to ensure correct selection under high competition.

Wilshire and McCarthy’s (2002) proposal is in line with the accumulating evidence for one or more mechanisms contained within the frontal executive system that operate to bias selection of linguistic and nonlinguistic representations when the demand is high. This idea, which dates back to Luria, 1966, Norman and Shallice, 1986, has been invigorated by contemporary neuroimaging and patient studies. In particular, Thompson-Schill and colleagues (Kan and Thompson-Schill, 2004a, Kan and Thompson-Schill, 2004b, Thompson-Schill, 2003, Thompson-Schill et al., 1997, Thompson-Schill et al., 1998) have amassed considerable evidence that the prefrontal cortex, specifically the posterior portion of the left inferior frontal gyrus, subserves such high-demand selection (and for evidence from an fMRI study of semantic interference in naming, Moss et al., in press). Moreover, recent patient studies link failure of controlled selection to paucity of propositional speech, a hallmark of anterior aphasia (Robinson et al., 1998, Robinson et al., 2005).

In summary, previous accounts of the blocking effect in nonaphasic and aphasic speakers present us with three theoretical positions to consider. The “too much inhibition” and “too much excitation” accounts propose alternative causes for the interference generated by blocking, while the executive selection account addresses the system’s response to such interference and is compatible with either causal account.

It is notoriously difficult to decide between excitatory and inhibitory accounts of semantic interference, as their predicted impact is the same whether one looks at naming latencies (longer with more excited competitors or greater target inhibition) or error rates (higher with more excited competitors or greater target inhibition) (see discussions in Forde and Humphreys, 1997, Wheeldon and Monsell, 1994). The examination of error types can be more revealing, but nonaphasic speakers normally produce too few errors to support such analyses. Requiring them to name pictures quickly is one way around this problem; Vitkovitch and colleagues’ use of speeded naming in several experiments provides the strongest evidence to date for “too much excitation” (Vitkovitch and Humphreys, 1991, Vitkovitch et al., 1993, Vitkovitch et al., 1996, Vitkovitch et al., 2001). However, speeded naming may not be sensitive to inhibitory effects that take time to develop or dissipate rapidly between trials (e.g. Neill & Westberry, 1987). Aphasic speakers generate sufficient errors even when naming at their natural pace. Moreover, the blocked-cyclic paradigm affords the potential for inhibitory processes that are normally transient to grow and build across cyclic repetitions (for empirical evidence of transient inhibitory effects with semantic blocking, see Campbell and Clark, 1989, Vitkovitch et al., 1996, Vitkovitch et al., 2001). For these reasons, we believe that the analysis of aphasic errors in blocked-cyclic naming presents the ideal circumstance for testing differential predictions of excitation and inhibition accounts.

The primary objective of this study was to investigate the three types of account using evidence from aphasic error rates and error types. The analysis of error rates addresses two questions: (1) Is a naming decrement in the homogeneous compared to mixed condition of blocked-cyclic naming a statistically reliable effect in the averaged performance of a mixed group of anterior and posterior aphasics?; (2) Is the blocking effect especially acute in a subgroup with nonfluent Broca’s aphasia—a classic anterior presentation—compared to fluent patients who match up closely with the Brocas on standard naming tests and other relevant language measures? Evidence for an exaggerated effect in the Broca group despite close matching would support the proposal that an anterior-based selection mechanism plays a role in the blocking effect.

The analysis of error types tests competing predictions from the “too much excitation” vs. “too much inhibition” accounts. The contrasting predictions are derived from Belke et al.’s (2005) excitatory, category node hypothesis and McCarthy and Kartsounis’s (2000) refractory inhibition hypothesis, but we will show that the findings also have implications for other excitatory and inhibitory accounts that make the same or similar predictions.

The organization of this paper is as follows: Part I presents two experiments: Experiment 1 investigates blocked-cyclic naming in older controls, with naming latency as the dependent variable. To date, the evidence that nonaphasic naming latencies are longer in homogeneous compared to mixed blocks comes from college-age participants. To verify that healthy older adults also show the effect, we first tested healthy participants, matched in age and education with our patient group. Experiment 2 investigates the same blocked-cyclic naming paradigm in persons with aphasia, with error rate as the dependent variable. Part II of the paper presents the analyses and theoretical interpretation of the error types that patients generated in Experiment 2. The General discussion considers what the collective findings say about the nature of the semantic blocking effect in blocked-cyclic naming and the implication of the patient findings for models of lexical retrieval in nonaphasic and aphasic speakers.