Attention and repetition priming in the verb generation task

Highlights

• Examined whether verb generation repetition priming is attention-dependent.

• Priming was reduced following divided attention and selective attention.

• DA did not differentially affect high-competition and low-competition priming.

• Results fit comfortably within the transfer-appropriate processing framework.

Abstract

Transfer-appropriate processing (TAP) and identification–production frameworks predict that repetition priming will be reduced by encoding-phase divided attention (DA) in implicit memory tasks that involve conceptual analysis of test stimuli and require responses that go beyond the identification of the test cue. This prediction was tested using the verb generation task. Verb generation priming was weakly affected by a number classification distracting task at encoding that impacted recognition, was affected more by a more demanding mental arithmetic task, and was abolished entirely by a selective attention manipulation. Priming originating largely from a process unique to the verb generation task was also found to be attention-sensitive. DA affected priming equivalently for high-competition and low-competition items, against the identification–production framework which predicts greater DA effects on priming in high-competition conditions. The results fit comfortably within the TAP framework.

Introduction

Cognitive psychologists have long known that division of attention during encoding has substantial and negative effects on direct or explicit tests of memory such as recall and recognition (e.g., Baddeley et al., 1984, Broadbent, 1958, Craik et al., 1996, Fernandes and Moscovitch, 2000, Murdock, 1965; for a review, see Mulligan, 2008). However, investigations of divided attention effects on indirect or implicit memory tests are comparatively recent. Early studies reported that repetition priming, a manifestation of implicit memory, was unaffected by the same divided attention manipulations that reduced explicit memory performance (Parkin et al., 1990, Parkin and Russo, 1990). Moreover, studies of priming for information encoded when attention is severely limited have yielded statistically reliable priming effects despite nonexistent explicit memory. Such studies involve, for example, presenting to-be-remembered information very rapidly (e.g., 2 ms/item; Mandler, Nakamura, & Van Zandt, 1987), diverting attention completely from the to-be-remembered information (e.g., Eich, 1984), or presenting information while participants are under surgical anesthesia (e.g., Kihlstrom, Schacter, Cork, Hurt, & Behr, 1990). Together, these initial findings suggested that attention was necessary for explicit memory but not for implicit memory, and that priming was the result of automatic encoding processes that operated independently of attentional resources (e.g., Eich, 1984, Jacoby et al., 1989, Parkin et al., 1990, Szymanski and MacLeod, 1996).

More recently, researchers have documented clear effects of divided attention (DA) on priming, bringing into question the view that implicit memory operates independently of attention. For example, relative to full attention (FA), encoding-phase DA has been shown to reduce priming on many implicit tests such as category exemplar generation (Gabrieli et al., 1999, Light et al., 2000a, Mulligan and Hartman, 1996), word stem completion (Clarys et al., 2000, Gabrieli et al., 1999), word naming (Light & Prull, 1995), general knowledge retrieval and word association (Mulligan, 1998), and perceptual identification (Mulligan, 2003, Mulligan and Hornstein, 2000; see Spataro, Cestari, & Rossi-Arnaud, 2011, for a recent review and meta-analysis). To account for these effects, researchers have considered the nature of the distracting task undertaken during encoding as well as the processing demands of the implicit memory test. With respect to the nature of the distracting task, DA effects on priming appear to be most evident with difficult or demanding tasks, with little or no DA effect occurring with relatively easy tasks (Mulligan, 1997, Wolters and Prinsen, 1997). In general, the effects of DA on priming appear to be most marked when distracting task stimuli are presented simultaneously rather than asynchronously with the to-be-learned information at encoding, and when the distracting task requires frequent rather than occasional responses (the distractor-selection hypothesis; Mulligan, 2003, Mulligan and Hornstein, 2000). Nevertheless, some results defy explanation based on distractor task difficulty, because priming on several implicit tests such as category verification, lexical decision, and object decision appears relatively immune to DA even under the aforementioned conditions (e.g., Light et al., 2000a, Mulligan and Peterson, 2008, Soldan et al., 2008).

Alternatively, the effects of DA can be understood by considering the processing demands of the implicit memory task. The transfer-appropriate processing framework (TAP; Roediger & McDermott, 1993) distinguishes between perceptual and conceptual memory processes, and predicts that memory test performance will improve to the degree that processes invoked at retrieval match or overlap those that were used at encoding. Therefore, priming on implicit tests that require perceptual analysis of stimuli (perceptual priming) should benefit from perceptual but not conceptual processing at encoding, and priming on implicit tests that require conceptual or semantic analysis of stimuli (conceptual priming) should benefit from conceptual but not perceptual processing at encoding. Divided attention at encoding is assumed to restrict conceptual processing relative to FA, leading to the prediction that DA should affect conceptual but not perceptual priming. This prediction has been supported in many ways, for example in the studies cited earlier reporting DA effects on conceptual priming tasks of category exemplar generation, word association, and general knowledge tasks, and little or no DA effects on perceptual priming tasks of lexical decision and object decision. However, inconsistent with this prediction, clear DA effects on priming have been shown for perceptual priming using the perceptual identification task, as mentioned earlier, and no effect of DA on priming has been reported for conceptual priming using the category verification task. Thus, not all findings can be accommodated by the TAP framework.

In contrast, the identification–production framework (Gabrieli et al., 1999) distinguishes between identification and production priming tasks, either of which can be perceptual or conceptual in nature. Identification priming involves stimulus identification, classification, or verification of a stimulus attribute (e.g., semantic verification, lexical decision, object decision, and word naming), whereas production priming requires producing one of several acceptable responses from the test cue (e.g., category exemplar generation, word stem completion, and, according to some arguments, perceptual identification (Mulligan & Peterson, 2008)). Compared to identification tasks, production tasks are thought to involve more response competition among multiple plausible alternatives at the time of retrieval, and this competition must be resolved in order to select and produce a response to a given test cue (see also Vaidya et al., 1997). The resolution of competition is aided by FA during encoding. Under FA, response competition is minimized on later production tests because perceptual and conceptual information about the encoded item is readily available and can be accessed quickly from an array of competing responses. This facilitation of access to encoded information following FA leads to maximal priming. However, following DA, this information is not as readily available so response competition is increased, which in turn leads to priming decrements. Gabrieli et al. (1999) provided behavioral and neuropsychological evidence to support the idea that production priming was more sensitive to DA than identification priming. Nevertheless, the identification–production framework cannot readily account for the clear DA effect on the identification task of word naming (Light & Prull, 1995). Moreover, the DA effect on word fragment completion priming is not any greater for fragments that have multiple solutions—and thus involve more response competition—compared to fragments that have unique solutions (Spataro, Mulligan, & Rossi-Arnaud, 2010).

The present study focuses on the prediction that these two theoretical frameworks make with respect to conceptual priming in tasks that require response production. Both frameworks predict that such priming will be sensitive to DA at encoding, and this prediction has largely been supported using a single task, the category exemplar generation task, which is both conceptual in nature and requires response production (e.g., Gabrieli et al., 1999; Light, Prull, & Kennison, 2000; Mulligan & Hartman, 1996; but see Baqués, Sáiz, & Bowers, 2004, for an exception). However, very little is known about how DA affects priming in other conceptual production priming tasks. Studies of the general knowledge task reported that priming was reduced by DA (Mulligan, 1998), and priming in a word association task was similarly reduced by DA in one study (Mulligan, 1998) but not in another (Koriat & Feuerstein, 1976). All of the aforementioned studies used tasks in which priming is expressed in terms of a change in accuracy; nothing is known about how DA affects conceptual production priming that is expressed as a change in reaction time (RT). The dependent measure of RT versus accuracy is an important consideration, for DA effects are often greater for accuracy measures of priming than for RT measures (Light et al., 2000b, Spataro et al., 2011).

The present investigation used the verb generation task to test the prediction that conceptual production priming will be affected by DA at encoding. In this task, participants encode nouns incidentally by generating an appropriate verb to each noun (e.g., pen → write). Later, the verb generation task is repeated with encoded nouns intermixed with new nouns, and priming is revealed by faster RTs to encoded nouns relative to new nouns (Raichle et al., 1994, Seger et al., 1999). Verb generation priming is not thought to reflect explicit memory because the magnitude of priming does not differ between amnesic patients and age-matched controls (Seger, Rabin, Zarella, & Gabrieli, 1997) or between young and older adult groups (Prull, 2004, Prull, 2010). Verb generation priming is considered to be conceptual in nature for several reasons and should therefore be affected by DA according to TAP. For instance, the task itself requires analysis of the noun's conceptual properties rather than perceptual analysis of the noun's appearance or form. Further, verb generation priming is sensitive to levels of processing effects (Seger et al., 1999, Exp. 5) and insensitive to stimulus modality changes from encoding to retrieval (Thompson-Schill & Kan, 2001, Exp. 2), which are hallmarks of many forms of conceptual priming (Roediger & McDermott, 1993). Verb generation is also clearly a production task because the task cannot be completed by merely identifying the noun stimulus, but requires that participants search and select one of many possible legitimate responses. Consequently, the identification–production framework predicts sensitivity of verb generation priming to DA. Finally, verb generation priming is expressed as a change in RT rather than accuracy. Finding that verb generation priming is sensitive to DA would be consistent with TAP and identification–production frameworks, but if such priming is not found to be sensitive to DA, that finding would be difficult to accommodate under either framework.

Although both frameworks predict DA effects on verb generation priming, only the identification–production account predicts further sensitivity of priming to DA with increased response competition at test. It is possible to manipulate response competition within a single production priming test, for example by presenting word stems, word fragments, or general knowledge questions that converge on only one possible solution (low competition) or can be answered with many possibilities (high competition; Barnhardt, 2004, Geraci and Hamilton, 2009, Spataro et al., 2010). In the verb generation task, response competition can be manipulated by presenting nouns that have a single dominant verb response (low competition, e.g., pen) or have multiple candidate verb responses (high competition, e.g., purse; Thompson-Schill, D'Esposito, Aguirre, & Farah, 1997). When the task is used as an implicit memory task, response competition may operate in the following way: Participants given low competition nouns at encoding (e.g., pen) are highly likely to produce the dominant verb (write), and that primed verb response can be rapidly produced again at test because it encounters little competition from few other infrequent and unprimed verb responses. In contrast, high competition nouns presented at encoding (e.g., purse) may elicit any number of verb responses (e.g., carry, hold, open, spend, steal), many of which are roughly equally probable. The primed response can be rapidly produced again at test but doing so requires resolving competition between that response and several other unprimed but readily available possibilities. The TAP framework makes no distinction between high and low competition test conditions and so predicts equivalent DA effects across conditions. However, the identification–production framework presumes that priming in high competition conditions is more attention-dependent than low-competition conditions, so finding greater DA effects in verb generation priming for high competition nouns would favor the identification–production account.