Abstract
This article reviews the Index of Linguistic Insecurity (ILI) as carried out for New York City by Labov and for Winnipeg by Owens and Baker and compares both to surveys done in southeastern Michigan in 2005, 2006, and 2007. The Michigan results, in spite of the apparent linguistic security there determined in earlier studies, reveal an even stronger insecurity than that reported for New York City. The article concludes that regional security does not imply personal security and shows a predominance of prescriptivist norms for the items surveyed in the Michigan studies as well as a higher incidence of insecurity for items that have poorly established community norms. The idea of linguistic insecurity as established in earlier studies is challenged by proposing instead a principle based on one’s fear of a personal inability to carry out a linguistic task.
Keywords
What is linguistic insecurity?
Speakers’ feeling that the variety they use is somehow inferior, ugly or bad. (Meyerhoff 2006:292)
Who has it?
Those who adopt a standard of correctness which is imposed from without, and from beyond the group which helped form their native speech pattern, are bound to show signs of linguistic insecurity. (Labov 2006:318)
Where does it come from?
The belief that “the language of a socially subordinate group is linguistically deficient” (Wolfram & Schilling-Estes 2006:398) and the corresponding “bias toward an abstracted, idealized, homogenous spoken language which is imposed and maintained by dominant bloc institutions and which names as its model the written language, but which is drawn primarily from the spoken language of the upper middle class” (Lippi-Green 1997:64).
How have we studied it?
The Index of Linguistic Insecurity (ILI), surely the most direct measure of this phenomenon, was devised by William Labov and calculated for New York City (NYC) in his 1966 study; it concluded that New Yorkers, especially lower-middle-class New Yorkers, were very insecure. A comparable investigation was completed roughly twenty years later in Winnipeg (WPEG) and found that Winnipeggers were more secure than New Yorkers (Owens & Baker 1984) but that the social class distribution was roughly the same. About twenty more years later (2005, 2006, 2007), undergraduates at a large southeastern Michigan (USA) university were given comparable tasks.
The results of all those studies are reported here and compared. 1 Although all these studies used alternative pronunciations to uncover speaker self-assurance, I will further show that the prescriptive status of the test items, the phonological/phonetic facts involved, and the surrounding language ideologies must also be taken into consideration for a reliable assessment. Finally, I reconsider the notion of “insecurity” itself.
Background
In the NYC ILI Labov (1966; the 2006 edition is cited henceforth) asked respondents which of two forms was correct and which they used (319), for example, [ɔftən] versus [ɔfn̩] (often). He tallied as cases of linguistic insecurity those cases in which a respondent said they did not personally use the one they judged correct (319). In his analysis, Labov regarded a 0 score as “no insecurity,” 1 to 2 as “mild,” 3-7 as “moderate,” and 8 and above as “heavy.” Table 1 lists the pairs that were presented and their alternative pronunciations (421).
Alternative Pronunciations of the Eighteen Items Presented in Labov’s (2006:421) Index of Linguistic Insecurity (ILI).
These items do not replicate Labov’s work on production, as in his Self Evaluation Test (SV), an alternative means of determining what he called “latent insecurity” (2006:319, and see below); in most cases, they also do not address specific elements of NYC speech, with the possible exception of the voicing concerns in because and Joseph and the [h] loss in humorous. They are instead mostly a collection of prescriptivist shibboleths, and a web search shows that avenue, February, catch, diapers, often, tomato, vase, arctic, aunt, because, and the h-loss in such items as humorous are still matters that should concern those who do not want, as one site advised, to use “mispronunciations that make you sound stupid” (Bowers 2008).
As in Labov’s studies of actual use, which showed that the lower middle class was most likely to hypercorrect, often using more of the local prestige norms than even the upper middle class in certain stylistic contexts (Labov 2006:309), the results of this survey indicate that the lower middle class is the one most beset with insecurity; 68 percent said they used three or more incorrect forms (i.e., moderate to heavy insecurity); in contrast, only 31 percent of lower-working-, 29 percent of upper-working-, and 10 percent of upper-middle-class respondents had such high levels.
Labov (2006:321) also reports average scores for women (3.6) and men (2.1), confirming the hypothesis based on his usage data that women would show greater linguistic insecurity. He also reports (321) the highest insecurity for Italians (3.8) and the lowest for African Americans (1.3), with Jews in a middle position (2.4), but he is unable to account for that pattern, and ethnicity will not be considered further here. Reporting on interviews with local speakers, Labov notes that “the term ‘linguistic self-hatred’ is not too extreme to refer to the situation which emerges from the interviews” (329-330); he concludes as follows:
This chapter developed one of the primary elements in the underlying structure of the New York City speech community: a profound linguistic insecurity. The lower middle class is the most seriously affected by this tendency, but all classes show this trait to a greater or lesser degree. (Labov 2006:322)
Owens and Baker (1984) replicated the ILI in WPEG, and supplemented it with a Canadian Index of Linguistic Insecurity (CILI), an additional twenty-two items selected from an earlier study of variable pronunciations in Canada (Scargill 1974). Their additional CILI items are listed in Table 2.
The Twenty-Two Items of the Canadian Index of Linguistic Insecurity (CILI) and Their Alternative Pronunciations (Owens & Baker 1984:343).
Owens and Baker (1984) justified their use of the ILI by stating that the United States has an “overpowering cultural influence . . . on Canadians,” and that, therefore, “Canadians are part of the North American linguistic community and can identify the class distinctions associated with the variable pronunciations of Labov’s list” (340). Many of the additional items on the CILI appear to be based on the encroachment of U.S. norms (e.g., lever, either). Owens and Baker changed the numeric range of the four levels of insecurity for both tests, but only slightly. The distribution of scores for the two WPEG tests is positively correlated; 71 percent of the respondents belonged to the same ILI and CILI category (Owens & Baker 1984:346). The major contrast with Labov’s (2006) findings is that WPEGers are less insecure than New Yorkers, who have 37 percent moderate and heavy insecurity combined compared to WPEGers 31 percent and 30 % WPEGers have 69 percent and 70 percent no and mild insecurity combined compared to New Yorkers’ 63 percent. Results for the WPEG ILI and the CILI, along with those for Labov’s ILI, are shown in Table 3.
Percentages of New Yorkers (NYC) and Winnipeggers (WPEG) Who Fit into the Four Levels of Insecurity of the WPEG Index of Linguistic Insecurity (ILI) and Canadian Index of Linguistic Insecurity (CILI) Tests.
Owens and Baker (1984) also confirm the considerable contribution to insecurity by the lower middle class, although consideration of the upper-middle-class pattern, which has only five respondents in the Canadian and ten in NYC, should be made with caution. This study will not deal with class distinctions, and the details of those findings are not reported here.
WPEG results also show that female insecurity scores are higher: WPEG ILI female = 2.23, male = 1.40; CILI female = 2.69, male = 1.72. The CILI scores are weighted by 0.818 to account for the different number of items, but the sex differences in both WPEG studies fail to reach statistical significance. Owens and Baker also display the contribution of individual items to both indexes. Tables 4 and 5 show these figures.
Itemized Percentages of Use, Based on Left and Right “Alternatives” and, in the Last Column, the Claim That Personal Use Was Incorrect, for the Eighteen Winnipeg (WPEG) Index of Linguistic Insecurity (ILI) Items (derived from Owens & Baker 1984:342 [Table 2] and 345 [Table 3]).
Itemized Percentages of Use, Based on Left and Right “Alternatives” and, in the Last Column, the Claim That Personal Use Was Incorrect, for the Twenty-Two Canadian Index of Linguistic Insecurity (CILI) Items (derived from Owens & Baker 1984:343, 345 [Table 4]).
The range of admitted incorrect scores for both WPEG studies is from 1.3 percent (length, ration, and cot) to 25 percent (February). If this range (23.7) is cut into four equal groups (intervals of 5.93), the following scale for each item’s contribution to insecurity results: 1-6 = weak, 7-12 = mild, 13-18 = moderate, and 19-25 = heavy. Table 6 shows the items grouped in this way.
Winnipeg (WPEG) Index of Linguistic Insecurity (ILI) and Canadian Index of Linguistic Insecurity (CILI) Items Assigned Four Insecurity Categories (derived from Owens & Baker 1984:345, Tables 3 and 4).
Owens and Baker (1984) conclude that spelling pronunciations are important contributors, although this cannot be directly derived from their data since their “admitted incorrect” category, like Labov’s, does not indicate which alternative this admission was a response to. For example, one might conclude, with regard to often, that the 22.5 percent (18) who said they did not use the correct form were mostly ones from the 44 percent (35) who said they used [ɔfn̩]. A similar interpretation might apply to February, diapers, arctic, and caramel, and perhaps whine, the last four all members of the moderate insecurity category. On the other hand, if Canadian usage assessments done some time ago are still valid, the spelling pronunciation of almond is dispreferred (e.g., Warkentyne 1971:198).
The remaining heavy and moderate items seem to focus on conservative Canadian (often English English) norms. Three (new, tune, and student) have to do with the presence of [j], a form almost completely gone from U.S. English (Labov, Ash & Boberg. 2006:54). Four—missile, schedule, progress, and butter—show Canadian and English English norms that differ from U.S. ones, but the source of the insecurity is not always clear. Two—progress and butter—show considerable variation in use (56 percent for the [ʌʊ] and [t] forms), and, as will be shown below, such divided usage can trigger insecurity in either direction, although it seems likely that the greatest insecurity would have been shown by those who said they used the [ɑ] and flapped forms respectively and regarded the more English English ones correct. On the other hand, 80 percent (64) of the respondents said they used [mɪsl̩] as opposed to English English [mɪsɑɪl], and 71 percent (57) indicated use of the U.S. [sk] form of schedule, reflecting much less divided usage than for progress and butter, but the interpretation may be the same—moderate or heavy insecurity on the basis of a failure to use the Canadian (or English English) conservative norm.
Owens and Baker (1984:343) also note a trend toward U.S. usage in the pronunciation of vase as [veɪz], but they are surely wrong. The most common U.S. pronunciation is [veɪs], and the alternatives presented by Labov (2006) were [veɪs] and [vɑz] (see Table 1). In this case, the source of insecurity would appear to result from the tension between the Canadian form [veɪz] and the English English form [vɑz].
An interesting high insecurity item is bury. The merger of stressed /ε/ and /ʌ/ (to /ʌ/) before intervocalic /r/ is a feature of Philadelphia-region U.S. English (Labov, Ash & Boberg 2006:54, 56-57), but there it is true of all /εr/ words, regardless of the orthography (i.e., it is as true of merry as it is of bury). This is not the case in Canada; this shift to /ʌ/ is limited to the lexical item bury and may, therefore, be listed with those that are spelling pronunciations (making all heavy insecurity a result of this fact). Owens and Baker (1984) do not comment on this form, and only 27 percent (22) attested that they used [bʌɹi], but the insecurity percentage is 21.3 (17), third highest for both scales. This pronunciation may have become a stereotype among Canadians, perhaps indicating old-fashioned and rural speech (J. K. Chambers, personal communication, 2010).
The Michigan Studies
Surveys done in 2005, 2006, and 2007 in Michigan, more than twenty years after the WPEG studies, used some items from the ILI and CILI, and a few new ones. Table 7 lists the items and their alternatives; the lists will be referred to as the 2005, 2006, and 2007 Michigan Index of Linguistic Insecurity (MILI).
Items Presented in the Michigan Index of Linguistic Insecurity (MILI) and the Alternative Pronunciations (respondent Ns are 2005 = 72, 2006 = 121, and 2007 = 766; total N = 959).
The 2007 form contained twenty-one items from the ILI and CILI. Deleted from the ILI were some stereotypes (e.g., vase) and from the CILI ones reflecting concerns limited to Canadian English (e.g., lieutenant). Two additional items used in 2005 looked at a Michigan shibboleth ([mεlk] for milk) 2 and at one item intended to assess sensitivity to the Northern Cities Shift (NCS)—the raising and fronting of /æ/ to [ε] ([bænd/bεnd]). The 2006 form contained all those and four additional NCS items: beg, to reflect the backing of /ε/ to [ʌ]; caught to reflect the lowering and fronting of /ɔ/ in the direction of [ɑ]; net, to reflect the lowering of /ε/ to [æ]; and cot, to reflect the fronting of /ɑ/ in the direction of [æ]. This last item is also included in the CILI, but there it contrasts [ɑ] and [ɒ], the latter an English English variant (see Table 2). Four other minor modifications of the targeted alternatives should be noted:
1) In the ILI the opposition [kεtʃəp/kætʃəp] (vowel only) was studied; in the MILI it is [kεtʃəp/kætsəp] (vowel and consonant).
2) In the CILI the opposition [kæɹəml̩/kæɹml̩] (number of syllables only) was studied; in the MILI it is [kεɹəml̩/kɑɹəml̩] (vowel only).
3) In the CILI the opposition [kəngɹætʃʊleɪt/kəngɹædʒʊleɪt] ([tʃ] versus [dʒ], voicing of affricate only) was studied; in the MILI it is [kəngɹætjuleɪt/kəngɹædʒuleɪt] ([tju] versus [dʒu], voicing and affrication; it is also the case that in the CILI the vowel is assumed to be [ʊ] but in the MILI that it is [u].
4) In the ILI for almond, the presence or absence of /l/ was studied; in the MILI it is the vowel alternative [ɑ] or [æ].
These very few changes should not interfere with overall comparisons of the ILI (NYC and WPEG) and the CILI with the MILI.
Michigan should be prime territory for this sort of investigation since all earlier work of both a quantitative (e.g., Preston 1996) and qualitative (e.g., Niedzielski & Preston 2003) nature has pointed to considerable security. Richard W. Bailey recalls this from his school days:
In my own case, the elite suburb of Detroit in which I grew up thrived on a certain smugness about its place in American life. My eleventh-grade class, I recall, was told of its good fortune because—unlike others—no one in our room spoke a “dialect.” (1973:385)
Even lower-status Detroit-area speakers agree: “If you have such a thing as called Standard English other than Textbook English it would probably be the language that you’re hearing right now” (Niedzielski & Preston 2003:99).
The methodology and demographics of the MILI differed from the ILI and CILI in the following ways:
First, instead of presenting auditory stimuli of the alternatives, they were presented in writing, with respellings or rhyming forms to indicate the choices. For example, ration was presented as follows:
ration A has an “a” that sounds like the “a” in “rat”
3
B has an “a” that sounds like the “a” in “rate”
Since student fieldworkers carried out the project, it was not advisable to have them pronounce the alternatives nor practical to equip them all with recorded versions to play for the respondents. It is impossible to assess exactly what influence this difference in mode of presentation had on the outcome; there was little or no reported objection from the respondents that they did not understand what was expected of them, and the failure of spelling to have any observable influence on their responses makes it seem that this procedure was a valid one. The entire test (illustrating the items used in all three MILI studies) is given in the appendix.
Second, although the respondents were all local (Michigan) young adults, there was no practical possibility of measuring their social status. All were undergraduate, university-enrolled persons, and only their sex was determined. Since items added to the 2005 and 2006 MILI studies targeted specific Michigan concerns, in the following comparisons I will use only the twenty-one items listed in Table 7 for the 2007 data, although I will pool the data for these twenty-one items from all three MILI studies (as the “CORE” data). I will comment on Michigan specific items separately below. Table 8 shows the overall results and results by sex (female = 484, male = 474) for insecurity for the twenty-one items. (Totals are reduced by items left unrated. Sex totals in the 2007 MILI are reduced by one respondent who did not indicate sex.)
Overall Scores and Scores by Sex (Twenty-One Michigan Index of Linguistic Insecurity [MILI] CORE items, 959 respondents).
As in both Labov (2006) and Owens and Baker (1984), the sex difference suggests greater insecurity for women. In the MILI CORE data, proportions of female versus male insecurity are significantly different (Yates-corrected χ2 = 13.508, 1 df, p = .0002; average female/male insecurity difference is significant in a two-tailed t test at p = .036).
The most straightforward way to make the overall results of the MILI more comparable to the earlier research efforts in NYC and WPEG is to calculate the number of individuals who scored 0, 1, 2, and so on for insecurity (i.e., those who said they did not use the form they selected as correct).
Tables 9 to 11 compare the MILI CORE results to the NYC and WPEG score distributions with the groupings used in those previous studies.
Comparison of the Michigan Index of Linguistic Insecurity (MILI) CORE to the New York (NYC) Index of Linguistic Insecurity (ILI) Based on the NYC Scale (χ2 = 6.11, 3 df, p = .106).
Comparison of the Michigan Index of Linguistic Insecurity (MILI) CORE to the Winnipeg (WPEG) Index of Linguistic Insecurity (ILI) Results Based on the WPEG ILI Scale (χ2 = 14.9, 3 df, p = .002).
Comparison of the Michigan Index of Linguistic Insecurity (MILI) CORE to the Winnipeg (WPEG) Canadian Index of Linguistic Insecurity (CILI) Results Based on the WPEG CILI Scale (χ2 = 17.9, 3 df, p = .0009).
The Michigan score distribution is not significantly different from the NYC one.
Tables 10 and 11 show that the MILI CORE results differ significantly from both WPEG ILI and WPEG CILI, and the direction of the difference is obvious: the Michiganders are more insecure.
How can Michiganders be more like the linguistically-prejudiced-against and self-incriminating New Yorkers than the linguistically-contented WPEGers. Before discussing why secure Michiganders respond this way, a few other comparisons can be made.
The MILI studies and both WPEG studies permit a more careful investigation of individual items. First, I will compare the items of the MILI CORE studies with those that overlap the two WPEG studies, looking at the rank order of intensity of insecurity for each item. Second, I will look at the MILI items independently to assess their degree of influence. Table 12 compares the percentage of insecurity for the twenty-one items shared by the MILI CORE and the WPEG ILI and CILI. Missing ranks represent items unique to MILI (see Table 13).
C = consonant difference; S = segment presence; V = vowel difference; *S = syllable count difference.
Items in Michigan Index of Linguistic Insecurity (MILI) 2005 and 2006 Not Shared with MILI 2007.
V = vowel difference.
The Spearman rank correlation of r = .4979 is significant (p = .02174), suggesting that there is general agreement between the MILI and WPEG measures on items that contribute most to linguistic insecurity. It is important to remember, however, that because, highest ranked in the MILI, was different in the two studies. The MILI looked at vowel variance ([ɔ] versus [ʌ]), while the WPEG investigators copied Labov’s (2006) focus on the voicing of the final consonant; the MILI vowel variation appears to have been more salient. The MILI looked at voicing and affrication in [kəngɹætjuleɪt] versus [kəngɹæʤuleɪt]; the WPEG study also looked at voicing but assumed affrication in both. The MILI looked at vowel differences in caramel ([ε] versus [ɑ]), while WPEG looked only at syllable loss (and assumed the vowel [æ] in both), 4 but, in spite of this different focus, the rankings were very similar (fourth and fifth). Finally, the MILI looked at the vowel variation in almond ([ɑ(l)mənd] versus [æ(l)mənd], but /l/ presence or absence was not studied, although that was the target in the WPEG. The /l/ factor caused rather more concern (MILI rank 19, WPEG rank 6).
Putting aside all these, although I need not have done so with caramel, the top-ranked items in both lists are February (second and first), arctic (third and fourth), and often (sixth and second). These rankings might seem to confirm Owens and Baker’s (1984) notion that spelling pronunciations (or avoidance of them) are a major source of insecurity among their respondents, particularly if almond is included (ranked sixth in the WPEG list). The top seven items in the WPEG studies are all marked “S” (“segment deletion”); the top four are deletions of consonants, and the fifth (caramel) is a vowel deletion that also results in syllable deletion. In contrast, the MILI list is not dominated by “S.” The top item is the vowel quality of because; the fourth-ranked item is caramel, studied in MILI for vowel quality only; the fifth-ranked item is congratulate, studied in MILI for both consonant voicing and segment type (stop+glide versus affricate), and the seventh-ranked item is catch, another case of vowel quality. Although the two lists correlate overall, spelling pronunciations seem to play a smaller role in MILI. Other items that were at rather different ranks (i.e., greater than fifth) were garage (MILI 10, WPEG 15.5), new (MILI 12, WPEG 3), apricot (MILI 13, WPEG 7.5), length (MILI 16, WPEG 20.5), 5 and diapers (MILI 20, WPEG 7.5).
Since the 2005 and 2006 MILI included a few items not used in the WPEG or NYC studies (milk, band, beg, caught, net) and one item (cot) shared between the 2005 and 2006 MILIs but only the CILI, it is possible that those scores might have contributed to the overall scores in some important way. They did, but in general it was to greater security.
Table 13 shows the results for these six items and calculates their ranks on the basis of where they would have fallen in the ranking of shared items shown in Table 12. Only the [mεlk] pronunciation of milk, a local stereotype, enters the ranking of items above the very bottom tier, and only Michigander security in their pronunciation of humorous (i.e., without /h/ deletion), ranks below any of these six items. Since, excluding milk, they were all intended to survey insecurity associated with NCS pronunciations, it is no surprise that they failed. Neither Preston (1997), with written stimuli, nor Niedzielski (1999), with auditory stimuli, were able to awaken recognition of the NCS in their respondents. The NCS is clearly change from below conscious awareness (e.g., Labov 2001:285), although this does not mean that sensitivity to its norms might not surface in certain sorts of investigations; these NCS items had to be represented in respellings that really did not capture the phonetic details of the shift, and these items were perhaps not successfully investigated in this mode.
Since, unlike the NYC and WPEG studies, the MILI studies correlate insecurity with which variant of an item was called correct, I can investigate those tendencies in more detail. In Table 14, column 2 (“Norm”) shows which form was identified as correct by the majority of the respondents and column 3 the form not so chosen (i.e., the “Non-norm”). Column 4 (“Norm Agreement”) shows the percent of respondent agreement on the norm in column 2, the basis of the ordering of this table. The percentage of overall insecurity for each item is shown in column 5, and column 6 shows the percentage of respondents who were insecure because they did not use the norm form in column 2. By inference, the remaining percentage (not shown here) identifies those who said the non-norm was correct and were insecure because of their use of the norm. In the first row, for example, 92.56 percent of the respondents (column 4) believed that the [kɑt] form of cot was correct, but fewer than 1 percent (0.83) were insecure with regard to this item overall (column 5). Of that small proportion who were insecure, however, all the insecurity arose from a failure to use the non-norm as indicated by the 0.00 percent who were insecure because of a failure to use the norm, the basis for the calculation of column 6. The situation is different for a case like humorous. Even though this item shows a comparable level of norm agreement (92.07 percent) to cot, there is a somewhat higher degree of insecurity overall (2.92 percent), and both pronunciations seem to contribute to that insecurity. Just over half (53.57 percent) of those respondents reporting insecurity were insecure because they failed to use the norm, and so the rest of the insecurity stems from failing to use the non-norm. In what follows, I will refer to these results as “norm agreement” (i.e., the degree of agreement on the majority variant), “insecurity,” (the overall percentage of insecurity caused by this item), and “norm insecurity” (i.e., the degree to which insecurity was caused by a failure to use the norm).
All Michigan Index of Linguistic Insecurity (MILI) Items, Showing the Percentage of Respondents Who Said That the “Norm” Alternative Was Correct (column 4), the Percentage of Overall Insecurity (column 5), and the Percentage whose Insecurity Was Based on Their Failure to Use the “Norm” (column 6).
Prescriptivists will be sorry to learn that Michiganders prefer the bisyllabic pronunciation of diapers; the form [kεʧəp] also carries the day over [kætsəp], which some respondents said sounded “prissy” or completely unfamiliar. It is surprising, however, in a task that so overtly probes notions of correctness, to see the forms [kεtʃ], [gəɹɑdʒ], [kəngɹæʤuleɪt], and [fεbjuεɹi] preferred as community norms, although none of them reaches even the 70 percent level of norm agreement. In contrast to February, the forms [ɔftn̩̩] and [ɑɹktɪk] showed a preference for the spelling pronunciations, but the slight preference for [bɪkʌz] goes against this trend, although spelling pronunciation with regard to vowels is harder to analyze. In general, these Michigan respondents display considerable variability in their norm agreement, with five cases even below the 60 percent level (congratulate, February, caramel, because, caught). They are also somewhat mixed in their norm bases; in arctic they agree with older, prescriptivist standards, but in many cases, their norm preferences point to newer ones (garage, February, because, congratulate, diapers, catch).
Figure 1 shows a scatterplot of norm agreement and insecurity, and, not surprisingly, there is an inverse correlation between the two: the greater the norm agreement, the less insecurity. Item 2 for example (humorous) has over 90 percent agreement that the norm is [hj], not [j], and fewer than 3 percent of the respondents show any insecurity for this item. Item 26, however, shows that only 53.55 percent agree that the norm is [bɪkʌz], and it is the largest contributor to insecurity (36.64 percent).
Scatterplot of norm agreement and insecurity (Pearson product–moment correlation r = –.7627929, 25 df, p < .01).
Item 27 (caught) deserves mention since it is considerably outside the trend line. Only slightly more than 50 percent of the respondents agree that [kɔt] is the norm, but, unlike all the other poorly-agreed-on norm items, it contributes almost nothing to insecurity. The answer to this anomaly lies, I believe, in its perceptual position in the NCS (and perhaps English in general). One standard story of the NCS is that /æ/ (hat) has fronted and raised, leaving behind a space into which /ɑ/ (hot) has moved; this leaves another space, and, in drag-chain fashion, /ɔ/ moves in (Labov, Ash & Boberg 2006:190-191). In a presentation of single-word items for phoneme recognition to young, European American southeastern Michigan respondents who could be expected to be advanced in the shift, Preston (2010a:249) shows that, with the exception of an /ε/ vowel realized as [ʌ], the respondents were least accurate in recognizing shifted /ɔ/; in 142 opportunities to identify it, exactly half (71) were misclassified as /ɑ/. The vowel in cot, however, is at the same end of the insecurity scale but the opposite end of the norm scale, suggesting that the NCS influence on perception of /ɑ/ was nil, and, indeed, Preston (2010a:249) reports on only thirteen mishearings of shifted /ɑ/ as /æ/, a further indication that these items are below the level of local speaker awareness and do not play a role in their linguistic insecurity. The historical status of the /ɔ/ vowel is undoubtedly a major factor in this confusion, and other studies (e.g., Peterson & Barney 1950) show difficulty in this territory as well, an issue that has traditionally been noted with regard to phoneme boundaries in the low-back area. As noted above, however, respelling is not the optimal way to measure sensitivity to these NCS items.
At the top of the norm agreement group (column 4 in Table 14), I have already discussed cot, a potential for influence by the NCS that was not realized. The next item, humorous, showed strong agreement on the [hj] onset as the local norm, and the alternative pronunciation was perhaps not familiar to many respondents, although the loss of [h] in some local words (e.g., Huron River) is common among older, rural, and less well-educated speakers. The [ε] form of beg is clearly the norm, and it caused little insecurity, another failure of a NCS item to influence these judgments, as was net a little farther down the scale (Item 8).
The [mɪlk] variant of milk is also the community norm but triggered insecurity in only 7.33 percent of respondents. Since the [mεlk] pronunciation is a local negative stereotype, it is not surprising that very few respondents reported using it, but that all who did were insecure. That is, of the few (7.33 percent) who exhibited any insecurity, all were insecure because they used [mεlk] (the 100 percent of column 6, Table 14). I have already mentioned that Michiganders prefer the bisyllabic pronunciation of diapers, and it also rouses little insecurity; [kεʧəp] is the same.
The items at the 80 percent to 89 percent level of norm agreement (lever, net, ration, band, apricot, length, and escalator) are, like those at the higher level, ones that caused little insecurity, and even at the 70 percent to 79 percent level of norm agreement, the items anti-, new, and almond are not very different from those listed just above, although anti- approaches 10 percent, the highest insecurity rate of any item yet commented on.
Things are very different at the next level down (60 percent to 69 percent norm agreement); the lowest norm agreement ranked item in this group, garage, provoked 12.51 percent insecurity. Even more dramatically, often triggered 20.46 percent insecurity (sixth highest in insecurity overall). These items, however, confirm the fact that insecurity is linked to uncertainty. For example, although the majority of insecure respondents agreed that, for example, the [ɹɑʊt] pronunciation for route was correct, we may infer from the norm insecurity scale of Table 14 (column 6) that about 60 percent of those who were insecure felt that [ɹut] was the correct form and that they did not use it. In other words, they were insecure in spite of their use of the norm—not a well-agreed on one.
At the bottom of the norm agreement scale (50 percent to 59 percent in Table 14), the leader in insecurity is because (36.64 percent). Although slightly over half the respondents identified the [ʌ] form as correct, the community norm was not the cause of the considerable insecurity. The norm insecurity scale shows that only 10.26 percent of the respondents who were insecure based their insecurity on the community norm; the extensive because insecurity was overwhelmingly based on the failure to use the [bɪkɔz] pronunciation (89.74 percent), the form that lost the norm agreement contest.
To look at norm, insecurity, and source of insecurity simultaneously, I will use a three-way feature system for the details in Table 14: for norm, items from 52.07 (the low) to 72.315 will be classified as “low” and items from 72.316 to 92.56 (the maximum) as “high.” For overall insecurity, items from 0.00 (the minimum) to 18.32 will be classified as low and those from 18.33 to 36.64 (the maximum) will be classified as high. For “norm insecurity” (i.e., insecurity based on a respondent’s saying they do not use the norm), items from 0.00 to 49.99 will be marked –, and items from 50.00 to 100.00 will be marked +. This classification is shown in Table 15.
All MILI Items Classified as High or Low Overall Insecurity and Norm Agreement.
As Figure 1 suggests, there should be no high insecurity/high norm agreement items, and the upper-left box of Table 15 is empty, but there is considerable diversity of affect among the remaining words submitted in the MILI. I have little doubt that similar diversity would be found among the NYC words, and some of this diversity is documented for the WPEG studies in Tables 4 and 5.
The most productive territory is low insecurity and high norm agreement in the upper right of Table 15. The four plus (+) items are those that cause some insecurity, however little, through failure to use the community norm: [mεlk] is a local stereotype, an alveolar /n/ in length is nonstandard throughout the United States, and the loss of /h/ in humorous is not widespread in Michigan and may be lexical where it does occur. The little insecurity in the pronunciation of lever appears to come from a failure to use the community norm, and the pronunciation with [i] is not common at all in the area.
The larger minus (–) group in the same area is more interesting since the small amount of insecurity here arises in opposition to the norm (i.e., people are insecure because they say they use the community norm and believe the alternative is correct). All the items added to the MILI list to test any insecurity rising from NCS influence (cot, beg, net, and band) are in this territory—high norm agreement (on the nonshifted form as correct) and little or even no insecurity. This is consistent with the change-from-below status of the shift and with previous studies that indicate local unawareness of the shift’s existence (e.g., Preston 1997; Niedzielski 1999). The other items (escalator, ration, ketchup, apricot, diapers, anti-, new) are items for which a small number of respondents felt that the majority norm was incorrect but that they were users of it. Several of them involve their apparent regard for very conservative and/or spelling pronunciations (new, ketchup, diapers, anti-). The item net has low insecurity and high norm agreement.
The next most productive area is that for items that were low in norm agreement but high in promoting insecurity, the lower left of Table 15. Only two items, arctic and often, caused considerable insecurity based on the weak majority norms. More interesting, the –norm agreement items are all ones that caused considerable insecurity, but, again, the source of the insecurity was based on the opposite norm. Since there was weak agreement on the norms in this area, perhaps not so much should be made of the distinction, and one may point to the norm disagreement as the major source of the insecurity. Nevertheless, that the majority of those who were insecure based their insecurity on the more conservative pronunciations of these five words suggests that considerable prescriptivist norm pressure exists even among these younger speakers.
In short, the two most likely outcomes are the two most populated areas in Table 15. Items about which there is considerable norm agreement cause little insecurity (the upper right) and items about which there is little norm agreement cause greater insecurity (the lower left). What about the mysterious lower right?
These five items (caught, route, either, garage, almond) caused little insecurity even though there was weak norm agreement. This is a surprising result, for a good sociolinguistic rule of thumb is that when two forms are in competition, the variants will take on social meaning. These items apparently do not awaken divergent social meanings; the weak norm agreement alone does not indicate that, but coupled with the low level of insecurity assigned these items, we find that is the case. The exceptional case of caught has already been discussed.
For weak norm agreement items, some have social meaning and may give rise to insecurity, but others do not. In the first case, this means that some respondents might choose one form for more formal or monitored usage while others would choose the alternate, failing to yield the unidirectional pattern given as proof of the uniformity of a speech community’s norms (Labov 1972:183-259). In the second, some items appear to have socially neutral variants, a confusing case for sociolinguists.
To summarize, the items in the upper right of Table 15 are ones that result in little insecurity and show a decided preference for one form or the other as standard in the speech community. The few people who go against the norm for some items do not increase insecurity in any interesting way. This set includes both items whose alternatives are not consciously known in the speech community (e.g., such NCS items as cot) and others that are stereotypes (e.g., the [ε] pronunciation of milk).
The lower-left items in Table 15 all contribute to insecurity, perhaps mainly on the basis of norm disagreement in the speech community. Unlike the first group, these are all community stereotypes, and the focus of the stereotypicality is on correctness, and this appears to be the major source of insecurity in the MILI studies and, I suspect, the others as well.
The lower-right items of Table 15 are, with the exception of caught (discussed above), community stereotypes, but, unlike the lower-left group, the focus is on variability rather than correctness. Discussions of route, for example, do not stem from attempts to reach some decision about correctness.
Since no ILI studies investigate empirically any stylistic differences in the alternatives presented and I have no social status characteristics available for this study, I can go no further.
Conclusions
Surprisingly, Michiganders equal or even outstrip New Yorkers in their insecurity. Any prediction that respondents who lived in areas where they felt their local variety was correct would be linguistically secure was flawed. What went wrong? Perhaps such local correctness may impose considerable obligations to behave correctly as individuals. In their assessment of their own speech, particularly concerning items for which they have neither prescriptive nor community norm guidance, they display an insecurity as great as those New Yorkers, who find their home area nonstandard. Perhaps there are those who find both their own speech and their own region correct; Hartley’s (1999) work in Oregon suggests it might be such a place, but specific work on insecurity has not been done there.
I suggest, therefore, a more careful distinction among insecurities. There are those who find their region (or group) incorrect and apparently extend that to personal insecurity (e.g., NYC respondents); there are those who find their own region (or group) relatively correct and extend that to their personal security (e.g., WPEGers), but there are also those who find their own area correct (perhaps even considerably so) but may find their individual performances lacking, particularly when local norms do not guide them. These last may be those who have considerably more invested in what Lippi-Green (1997) calls the “standard language ideology,” the belief that some varieties are more correct than others. Michiganders certainly belong to this group, for they do not hesitate calling Michigan English the most correct in the United States when asked to rank varieties on this dimension (by U.S. states) (Preston 1996). On the other hand, they seem to have much less invested in other forms of prestige. When asked to rate the local area and the usually stigmatized U.S. South, Michiganders found the South to be superior to their own speech on such dimensions as “casual,” “friendly,” “down-to-earth,” and “polite” (Preston 1999:366). U.S. Southerners, however, show a much greater investment in such solidarity dimensions when asked where “pleasant” and “correct” English are spoken, unfailingly identifying their own locale as the site of the most pleasant speech (e.g., Preston 1996). I conclude that they are somewhat less invested in a standard language ideology, and a look at their linguistic insecurity would be very interesting but poses a problem.
All the ILI work reviewed here presupposes a unidimensionality based on what Trudgill (1972) would call overt prestige or the correctness norm basis for insecurity noted in the quotations at the beginning of this article. I believe we should be more careful in determining what specific ideological factors are at work and how they affect individual tasks, performances, and even individual items presented for evaluation as we carry out such research. I specifically urge distinguishing between regional (or group) and personal insecurity. In the earlier work on Michigan correctness and pleasantness, for example, African American Michiganders, whose own speech variety may be discriminated against, rated Michigan speech as even more correct than their European American counterparts did (Niedzielski & Preston 2003:75). They were obviously participating in the standard language ideology of the area, one that would have included disparagement of African American Vernacular English.
I am also concerned with the either–or division of both status (overt) and solidarity (covert) prestige. Let me comment on just one upshot of this concern—the superstandard.
Standard American English, in the informal sense, or the informal standard form of any language, must be distinguished not only from substandard forms but also from superstandard forms. There is general agreement about what forms of language are preferred above others within a language community, even when the preferred forms are not used. It is typical for people to be slightly schizophrenic about their use of language. They acknowledge that some aspects of their language use are not “correct”: they can tell you what the “correct” form is, but they never actually adopt it. At an emotional level, these admittedly correct forms are rejected by some speakers because they are too correct. These speakers do not adopt such forms and at unguarded moments will even make negative value judgments about speakers who use them, not because these forms are “bad English” or because the speakers who use them are considered uneducated, but because the forms are “too snooty” or “too high-falutin’.” (Wolfram & Fasold 1974:19)
Although I believe I have shown in both the WPEG and MILI surveys that Wolfram and Fasold (1974) exaggerate a bit when they say that “there is general agreement about what forms of language are preferred above others within a language community,” there is other evidence to support their notion of a superstandard. In many of the dialect map-drawing tasks assigned Michigan respondents in earlier research, results such as those shown in Figure 2 emerged.
Hand-drawn map of U.S. dialects by a Michigan respondent (Hartley & Preston 1999:224).
The home site (Michigan) has the character that those who would hope their correct but not too correct language would have—boring, but the entire Northeast, for this respondent, shows “
If this populist aspect of U.S. language regard is true, and I believe it is, perhaps there is no insecurity at all in a respondent who tells us that [vɑz] is correct but that they say [veɪs]. Perhaps many of the highly ranked MILI sources of insecurity (e.g., February, arctic, caramel) have this character; their fancier pronunciations are correct only in the superstandard sense, and their alternatives are normal, hardly a source of insecurity.
The underlying strategy of ILI-inspired research is not wrong if we take the view that sociocultural prescription surrounding language is a primary contributor to insecurity, and much social-psychological, sociolinguistic, ethnographic, language-ideological, folk-linguistic, and perceptual-dialectological research would confirm that. But I also believe that research methodologies that focus primarily on such niceties, as the above analysis shows all three studies have done, may miss opportunities for the discovery and cultural weighting of other kinds of linguistic misgivings. I prefer the following definition (at least for the “what is it” and “who has it” questions posed at the beginning):
Linguistic insecurity arises when one feels that they are not able to perform the linguistic job at hand.
From this perspective, insecurity ranges from lacking any skills at all in a foreign language all the way to slight misgivings that one did not properly assess a situation in terms of the face or identity that they meant to project in their contribution and therefore failed to match items from their repertoire to that instance of performance. This definition allows us to think about ways of assessing speakers’ worries about their abilities to perform in any way that would satisfy their identity construction of the moment. One may go on to imagine a long list of linguistic inabilities, not only those that are not a part of our repertoires but also imagined or real failures of our ability to select correctly from those that are solidly in our repertoires. Fear of public speaking is well known, but for those of us more interested in the daily stuff of language in public life, it is rather fear of speaking in public that might characterize linguistic insecurity.
To conclude, I believe that measures of linguistic security, however quick and dirty they may seem, form an important part of the investigation of language regard (e.g., Preston 2010b), but they are surely qualified by at least the following conditions:
They should not presuppose that regional (or group) (in)security will be reflected in personal (in)security. 6 Different speech communities (or subspeech communities) may reveal different patterns.
They need to be correlated with qualitative measures of regional or group security such as ethnographic interview techniques (e.g., Labov 2006:300-323; Niedzielski & Preston 2003:97-126). In fact, Labov’s category “stereotype” (e.g., 1972:248) can be determined in only this way.
Individual items submitted for judgment should be carefully accounted for in terms of the sort of (in)security each triggers (e.g., Graff, Labov & Harris 1986; Milroy & Preston 1999; Plichta & Preston 2005; Preston & Niedzielski 2010).
The manner of presentation (setting, task, etc.) will have an important influence on the regard assessments made by respondents, including the growing concern with explicit versus implicit judgments (e.g., Bassili & Brown 2005), which has already found particular expression in work on language attitudes to varieties of Danish (e.g., Kristiansen 2009) and phonological features of Texas speech (e.g., Koops, Gentry & Pantos 2008), although earlier research was also conducted in which such overtly conscious measures as the ILI were not used (e.g., Labov’s SV [2006:300-317], Trudgill’s [1972] study of covert prestige in male responses in self-evaluation, and Niedzielski’s [1999] study of self-awareness of the NCS among southeastern Michigan respondents).
Until recently, the majority of our investigations in quantitative sociolinguistics has been of phonetic, phonological, and morphophonological factors, but, however carefully contextualized, we may have missed in some cases the important contribution of the lexical items themselves and therefore the need for finer distinctions among the types of linguistic units that play a role in language variation and change. Perhaps we need not incorporate the dictum chaque mot a son histoire, but we would do well to pay attention to the status of the phonetic realization of the vocabulary in our sociolinguistic work. Recent work has already, for example, focused on status frequency in variation and change. Finally, I believe that, as in many areas of the social sciences, a variety of approaches will reveal a more complex but also more interpretable picture (Preston 2010b).
Footnotes
Appendix
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.