To what Extent do Speech Errors serve as Linguistic Evidence?

Inhaltsverzeichnis

General Introduction

1. General notions of speech production and error types
1.1 Theoretical and biological approaches – models of speech production
1.2 The First Law and historical change
1.3 Distinguishing errors and associated difficulties

2. Error types and inferred conclusions
2.1 Anticipation and Perseveration errors
2.2 Omission and Addition errors
2.3 Exchange and Substitution errors
2.4 Blends and Spoonerisms

3. Summary of presented ideas and conclusion

Bibliography

General Introduction

Theories about speech production and its underlying rules are of increasing interest for linguistic research and have been for many years already. As Booij puts it:

“(…) [T]he mental representation of morphological knowledge is a battle ground for different theories about the nature of linguistic rules (…).”[1]

Errors of speech play an important role in these theories, as do errors in reading and writing. Although latter error types deliver further evidence supporting the ideas presented in this paper, the considerations gathered will, in default of space, be restricted to slips of the tongue. This error type is by Boomer’s and Laver’s definition:

“(…) an involuntary deviation in performance from the speaker’s current phonological, grammatical or lexical intention.”[2]

Further distinctions will be made in respective chapters of this paper. Each error type will be illustrated by examples found in the appendixes of Fromkin’s “Speech Errors as Linguistic Evidence” (1973) and Cutler’s “Slips of the Tongue and Language Production” (1982). All of the presented examples will be indented and made up in the same way: the intended sentence, phrase or word is to be found on the left, the erroneous output follows after the symbol “ ”.

Where it is possible, personal observations and own examples are added.

The difficulties of collecting errors under both natural (i.e. during conversation etc) and non-natural (i.e. in a laboratory etc) conditions[3] will be compared as will be considered what makes a slip of the tongue regarded as false (i.e. what mechanisms, biological and linguistic, are behind the effect hearers regard as an error of speech)[4]. Closely connected with these problems are error frequencies and their linguistic implications which will be regarded as well.

Furthermore, models and theories of the rules underlying speech production, both early and current, will be compared and discussed. However, Freudian approaches will mostly be left out, as there is spread consensus that the corpus of speech errors collected by the Austrian psycho-analyst tends to be biased by his theories. It is nevertheless one of the earliest corpuses collected and delivers valuable psychological data. And though one could argue that every theoretical model is biased in a certain way[5], the mere linguistic, grammatical and biological aspects of producing speech errors will be of interest, not necessarily the psychological process taking place under this surface and being hardly detectable by statistical means.

Concluding, this paper tries to illuminate the importance of speech errors as a way of gaining insight into the mental lexicon and achieving a better understanding of how speech is produced, in short: to what extent speech error data can offer a window to the speaker’s mind.

1. General notions of speech production and error types

1.1 Theoretical and biological approaches – models of speech production

Cohen offers a three-staged model of speech production (based on Lashley’s groundwork); there is

“a) a plan, determining what is going to be said
b) a programme controlling the temporal ordering in accordance with rules
c) the actual performance (…)”[6]

The actual performance is considered as a biological process and some linguists assume that errors not only occur on lexical or semantical level (i.e. during planning and controlling) but also during speech itself. Laver argues, that “[s]peech (…) involves very fast, complex movements”[7] and therefore suggests the existence of automatically operating speech patterns. Such defaults would allow to direct attention to other segments of speech production, e.g. pronunciation. The defaults themselves are maybe not as thoroughly controlled as other segments. Alas, an error might lead to what is often referred to as tripping over one’s tongue.

Laver also points out how to actually define speech errors, i.e. how to judge when something is to be regarded as wrong. He does so by using the notion of speech as a programme and speech errors as a programme that differs from the plan the speaker intended.

Therefore, errors are either to be found on the lexical (words that don’t exist in the regarded language, see example 1 a)) or semantical level (words that do exist in the regarded language but are used in an inappropriate and / or wrong context, see example 1 b))[8].

1. a) didn’t bother me in the least didn’t bother me in the sleast

b) our dear old queen our queer old dean

This idea is proposed by Butterworth as well, who divides the mental lexicon into a semantic and phonological half, detecting the possible occurrence of errors in both and thus coming to similar conclusions as illustrated in 1 a) and b)[9]. Laver furthermore suggests reflecting on the term “normal speech” and claims that “normal speech is not necessarily error-free speech, and (…) errors are part (…) of normal speech”[10], making them a worthy field of linguistic research. However, fitting errors into models of speech production appears to be a highly theoretical approach, as most of the processes involved seem to function simultaneously. The problem here is that, of course, most models work as linear ones and therefore are forced to leave out important aspects, as Butterworth points out[11] (his own model tries to avoid this dilemma by assuming two parallel stages or halves in which errors might be created, as mentioned above). His assumptions offer a revision of older models and nourish the theory of the mental lexicon as a multi-dimensional network rather than a linear collection of entries; this is proposed by other linguists such as Fromkin[12], Nooteboom[13] and Booij[14] as well.

1.2 The First Law and historical change

In addition to Laver’s approaches, Wells proposes three further laws of speech errors, meaning statistically extremely high probabilities regarding the way errors appear. The second and the third law will not be examined closer as they consider special circumstances, under which errors might occur, in detail.

But the first law is of general importance:

“A slip of the tongue is practically always a phonetically possible noise.”[15]

(Meaning a phonologically possible cluster of components in the respective language considered, as is also proven by Fromkin[16]). Wells thereby indicates that speech errors do not appear randomly and can be predicted to a certain extent.

These patterns may change in the course of time, as the following example of historical change in the English language illustrates:

2. a) short lady short shlady

b) shut his mouth shmut his mouth

The errors created were once regarded as phonologically impossible. However, the adaption of Yiddish words (e.g. “shmuck”) obviously led to an adjustment of the First Law[17]. Still, it is to be considered as the basic framework in which speculations about potential errors can be made. It is therefore not devaluated in my eyes.

1.3 Distinguishing errors and associated difficulties

There are numerous ways of arranging and differentiating errors collected in linguistic research. Many linguists develop their own systems and sub-categorizations, making it hard to gather basic examples and general notions. Nevertheless, some cardinal errors can be detected and will be discussed here – further refinements will be made only if necessary in order to account for certain ideas. This is, again, due to lack of space.

Four large groups of speech errors will be assumed, based on theories presented in Cutler’s (et al.) “Slips of the Tongue in the London-Lund Corpus”[18]. These groups are:

1. Segment errors (involving phonemes / phonological features)
2. Syllable and morphological errors (involving grammatical morphemes)
3. Word errors (involving the substitution of either a semantically or phonetically similar word)
4. Errors concerning larger units (involving blends of phrases)

Each group will be divided into further error categories such as anticipation, Perseveration, omission, addition, exchange, substitution, and blend[19]. While the four major categories given above describe the level on which an error is found in a regarded structure, the latter seven refinements describe the actual appearance of these errors and will be examined more closely in the following chapter. Furthermore, conclusions that can be drawn from these examples will be discussed, taking the following principles, established by Cutler[20], into account:

1. Some errors are like abc, therefore…
2. More errors are like abc than like xyz, therefore…
3. No errors are like abc, therefore…

This distinction of arguments offers a comprehensible and traceable way of fortifying speech production theories and performance models with the data and examples discussed below.

[...]