Why do we say ‘salt and pepper’ but not ‘pepper and salt?’ In this paper, we address this question by investigating the degree of entrenchment of familiar and less familiar binomial expressions. To do this we use the paradigm ‘fast pairs’ developed by Caldwell-Harris and Morris (2008) and apply it for the first time to binomial expressions to examine the tendency to make reporting errors when binomial expressions are presented in fast succession. The binomials discussed are examples of formulaic language, defined by Wray (2005) as a sequence of words, which is, or appears to be stored and retrieved from memory at the time of use. Formulaic sequences are pervasive in language use and make up a large proportion of any discourse (Conklin & Schmitt, 2012). For formulaic language to be so widespread in everyday language it must be purposeful in communication. Previous research has largely disregarded a research question that is interested in the flexibility of formulaic language, specifically binomial expressions. Hence, in this paper, we investigate reporting errors in order perception to extend the programme of research in the School of Psychology on formulaic language.
The term ‘binomial’ was originally introduced by Malkiel to define the sequence of two words that are associated with the same form-class, have an identical level of syntactic hierarchy and are connected via a lexical link (Malkiel, 1959). Malkiel (1959) further introduced the cline of reversibility, operationalised in terms of the relative proportions of the corpus-attested frequency of preferred and non-preferred binomial sequences. A reversibility scale therefore exists, whereby some binomials are irreversible and others are completely reversible or prefer one order over the other to any degree (Mollin, 2014). This is determined using the British National Corpus (BNC), a 100-million-word corpus collected to represent spoken and written British English across a wide range of different linguistic sources (BNC Consortium, 2017). According to Malkiel (1959), a binomial expression is frozen or irreversible if the word order is almost exclusively fixed and only one of the two possible sequences occurs in the corpus, for example, the expression fish and chips is considered frozen as it never occurs as chips and fish, whilst short and long is fully reversible as it is encountered almost equally as long and short. The frequency of a binomial is then determined based on its level of occurrence in the corpora, for example, law and order is considered a high-frequency binomial due to its high level of occurrence in the corpora, whilst son and heir is considered low frequency. As such, back and forth is an example of a high frequency, frozen binomial, whilst willing and able is a low frequency, flexible binomial.
The study of frequency effects has its origin in the seminal psychological research of
Cattell (1886). Cattell was the first to demonstrate the word frequency effect, i.e. that
higher frequency words are recognized more quickly than lower frequency words
The study of frequency effects has its origin in the seminal psychological research of
Cattell (1886). Cattell was the first to demonstrate the word frequency effect, i.e. that
higher frequency words are recognized more quickly than lower frequency words
The study of frequency effects has its origin in the seminal psychological research of
Cattell (1886). Cattell was the first to demonstrate the word frequency effect, i.e. that
higher frequency words are recognized more quickly than lower frequency words
Langacker (1987) was one of the first researchers to assume that the frequency with which words occur in language influences how they are processed, suggesting that frequency of exposure correlates with the degree of entrenchment and further, that depth of entrenchment may lead to the unitisation or holistic representation of a phrase. Langacker (2008) later related entrenchment to the process of automatisation, whereby, a complex structure is mastered through repetition or rehearsal making it effectively automatic. Subsequently, high-frequency phrases become automatic, creating advantages for word recognition and production. Following Langacker (1987) this paper uses entrenchment to refer to the hypothetical mental processes of routinisation and over-learning resulting from repeated exposure to meaningful stimuli. Langacker’s (1987) position that formulaic language has a processing speed advantage has since been supported by a number of research approaches, with studies appearing to be congruent in indicating faster processing and better representation of more frequent multiword sequences compared to less frequent ones (Balota & Chumbley, 1984; Monsell, Doyle & Haggard, 1989; Rayner & Duffy, 1986).
However, recent eye-tracking work by Siyanova-Chanturia, Conklin and Schmitt (2011) was one of the first empirical studies to indicate this holistic storage of frequent multi-word items. In Siyanova-Chanturia et al. (2011) study, participants read sentences containing three-word binomials phrases that were identical in syntax but different in frequency, such as bride and groom, in their normal and reversed forms (groom and bride). They found that more frequently occurring formulaic sequences appeared to be more easily and quickly processed by native and non-native speakers than less frequent ones. Thus, supporting the view that frequently encountering a sequence of words contributes to its degree of entrenchment (Langacker, 1987; Schmitt & Carter, 2004). Regardless of frequency, the typical binomial phrases were also processed more quickly than their reversed forms (Schmitt & Carter, 2004), suggesting that the familiar order is holistically stored in the lexicon. Albeit, Siyanova-Chanturia et al. (2011) study had a very small group of respondents, therefore, whilst findings display the important role of frequency in language processing and emphasise entrenchment for phrasal configuration in memory, it cannot be taken to suggest that formulaic sequences are necessarily processed as holistic units.
Lohmann (2011) similarly discusses the holistic storage of binomials, however, relative to the concept of reversibility. Whilst no empirical evidence has been found to corroborate his assumptions, Lohmann (2011) theorised that less reversible binomials may possess their own presentation in the mental lexicon, having frozen, so that no active computation plays a part anymore in their production, thus, irreversible binomials become holistically stored. Whereas, reversible binomials are computed ad hoc, meaning the sequence of their elements is not fixed and therefore they are ordered according to lexical accessibility. However, it must be considered that Lohmann’s (2011) reliance on corpus data may reflect the production of an epiphenomenon (Leech, 1992; Mollin, 2014). Rather, empirical evidence on the processing of binomials is provided by Arcara et al. (2012). Through a neuropsychological investigation, Arcara et al. (2012) investigated the mental representation and processing of irreversible binomials in patients with neglect dyslexia. Neglect dyslexia compromises a patient’s ability to read stimuli on the left side of the visual shield and therefore, the researchers made the assumption that if patients with neglect dyslexia show a phrasal effect, this would signify that these phrases are represented as whole in the patient’s mental lexicon. The patholinguistic study revealed that people with left-sided neglect dyslexia read irreversible binomials better when their components were presented in the correct order, indicating that they processed irreversible binomials as lexical elements (Arcara et al., 2012). Whilst this study appears gives strong evidence for the holistic storage of binomials in the mental lexicon, language processing in neglect dyslexia remains fairly under-researched (Mollin, 2014). Hence, to draw meaningful conclusions here further research is needed.
Mollin (2014) was the first empirical research specifically into the factors governing the cline of reversibility of English binomials. In her corpus-based study, native and non-native speakers were asked to make intuitive judgments about the reversibility of selected binomials to test whether native speakers were able to replicate corpus data on reversibility in their responses (Mollin, 2014). Non-native speaker’s judgments about reversibility were found to diverge significantly more strongly from corpus values than native speaker’s judgments, suggesting that a lack of examples encountered in language can hamper the build of accurate intuitions (Mollin, 2014). This comparison supports the assumption that intuitions about reversibility are the result of exposure leaving traces in the mental lexicon. In this instance, we would expect proficient speakers of the English language to be aware of the degrees of reversibility due to their amalgamation exposure building their intuition on the relative frequencies of binomial sequences. Whilst, the meaningfulness of Mollin’s (2014) findings have been questioned due to subjectivity (Dabrowska, 2010), judgment data can provide a valuable form of insight when integrated with corpus methodology for the purpose of this study (Gilquin & Greis, 2009). This paper attempts to yield Mollin’s (2014) desirability for future research on the processing of more or less reversible binomials and their sequences in addition to frequency by varying the temporal order of words.
Temporal order judgment tasks are a well-established method of examining errors in perceiving order (Stolz, 1999). According to Reeves and Sperling (1986), the order in which we perceive events is in the order in which we believe they occurred. However, it is well known that rapid visual displays lead to errors in reporting order (Reeves & Sperling, 1986; Scarborough & Sternberg, 1967), with the close temporal proximity causing interference with the target identification task and produce an attentional blink (Raymond, Shapiro & Arnell, 1992). Despite being well-established, the first research into how familiar order influences perceptions of the temporal order of words has only recently carried out by Caldwell-Harris and Morris (2008). The researchers first introduced the fast pairs paradigm, that is, a visual word recognition paradigm for measuring entrenchment, top-down effects and subjective phenomenology to explore the discrepancies between perceived and actual order (Caldwell-Harris & Morris, 2008). Using this paradigm, pairs of co-occurring word sequences were briefly flashed on a computer screen, one word immediately after the other and participants were asked to report the words in the order that they saw them. Caldwell-Harris and Morris (2008) compared high-frequency collocations, with low-frequency collocations, with word pairs that were merely adjective and noun combinations and with random combinations. In line with Mozer (1983), when words that had a familiar order were detected, the familiar unit was likely to be perceived. Caldwell-Harris and Morris (2008) found that participants reversed word pairs that had a familiar order and denied purposely changing the order or guessing. According to Lupyan and Clark (2015), repeated exposure to language can also create an illusion of prediction rather than reality. Their framework on predictive coding provides a plausible mechanism for many of the reported effects of language on perception, explaining that the brain combines prior knowledge and expectations with incoming sensory information to conceive a perception that reflects its best available hypothesis (Lupyan & Clark, 2015). However, in the fast pairs study reversal errors also remained frequent even during longer duration trials, suggesting that observers have strong phenomenology of seeing word pairs in a familiar order despite their temporal occurrence. Thus, Caldwell-Harris & Morris’ (2008) fast pairs study supports the previously discussed findings, giving strong support for the assumption that frequently occurring phrases are mentally entrenched in memory (Langacker, 1987; Siyanova-Chanturia et al., 2011). The present study examines a similar tendency to make reporting errors using this paradigm, however, with binomial expressions.
Rationale
As has egressed, there is great capacity for research on the processing of binomials using the fast pairs paradigm (Caldwell-Harris & Morris, 2008). There is some debate amongst linguists about the factors that govern the sequence of elements in binomials (Malkiel, 1959: Mollin, 2014), however, previous empirical research on binomial frequency and reversibility in the mental lexicon is scarce. With the exception of Caldwell-Harris and Morris (2008), there has been no attempt to investigate the degree of entrenchment and reporting errors in word order within the domain of formulaic language. This paper, therefore, intends to extend this dearth of research on binomial frequency and reversibility by manipulating word order and using the fast pairs paradigm developed by Caldwell-Harris and Morris (2008) to investigate if individuals do experience mental entrenchment (Langacker, 1987). Common phrases such as ‘salt and pepper’ will be presented rapidly and briefly on a computer screen to examine how word frequency, reversibility and order affects adults’ perception of formulaic sequences. We also use a second, self-paced reversibility rating task, involving a 7-point scale to assess observer’s familiarity with phrases in attempt to provide additional evidence for the status or reversibility in the mental lexicon and to give an indication of the storage and processing of binomials and to further investigate corpus-based reversibility scores.
Hypothesis
Consistent with Wray’s (2005) definition of formulaic language, previous studies have found that participants make faster and more accurate judgments about the word order of binomial expressions when the normal word order is encountered and also when the expression has a higher frequency of occurrence in everyday language (Caldwell-Harris & Morris, 2008; Siyanova-Chanturia et al., 2011). On the basis of previous literature, it is therefore hypothesised that participants will make more reversal errors when the word order of binomials is reversed compared to when the phrase is presented in its ‘normal’ order. Frequent exposure to binomials appears to lead a particular form of a sequence becoming entrenched in the mental lexicon (Siyanova-Chanturia et al., 2011). Hence, it is predicted that participants will make more reversal errors for high-frequency phrases compared to low-frequency phrases and additionally, reversal errors will be more likely for irreversible binomials than for reversible binomials. The body of research reviewed also implies that individuals who have had sufficient exposure to a language appear to have representations for formulaic sequences (Mollin, 2014). Therefore, in the reversibility rating task, it is predicted that participants’ judgments will converge with corpus values.
